| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 310.1 | FWD: MCC Systems Report | ULYSSE::ROACH | TANSTAAFL ! | Fri Mar 29 1991 10:05 | 214 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 28-Mar-1991 05:22pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD20@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC Systems Report
MCC Systems Technology Report INTERNAL USE ONLY
*****************************
March 27, 1991 (180 lines)
Prepared by Bill Kuhlman and Ralph Cherubini, Austin, TX
This newsletter covers the work in the Object-Oriented and Distributed Systems
(OODS) lab within MCC's Advanced Computing Technology (ACT) Program and the
Deductive Computing Lab within the Software Technology Program (STP). Please
feel free to forward it within Digital. Address inquiries concerning Deductive
Computing projects to AUSTIN::CHERUBINI; OODS projects to AUSTIN::KUHLMAN.
Technical Seminar on ESP at ZKO - April 5, 1991
-----------------------------------------------
Kim Smith, with MCC's Experimental Systems lab for parallel processing, will
present a Technical Seminar Series talk Friday, April 5, from 10:00-12:00 a.m.,
in the Babbage auditorium, ZKO1-3, Spit Brook Rd., Nashua. Titled "Extensible
Software Platform" (ESP), the presentation will discuss the C++ object-oriented
paradigm as the enabling technology for ESP.
OODS (Carnot)
=============
Carnot visits Digital
---------------------
Phil Cannata, Director of the Carnot project, gave a TSS talk on "Enterprise-
Wide Information Management: The Carnot Project at MCC" on March 13 at ZKO.
It was attended by about 35 Digital engineers and managers. Following the
talk, meetings were held with members of Database Systems Group and the
Cambridge Research Lab. Digital is not currently a supporting member of the
Carnot project, and we need to decide shortly if we will participate.
Dr. Cannata will visit Digital's Japan Research And Development Center,
Yokohama, on April 3, 10:00 am to discuss the Carnot project.
Distributed Transaction Environment via Rosette
-----------------------------------------------
In the past month, a Service Agent abstraction was incorporated in Rosette to
support implementing the semantics of OSI services. A number of residual
errors in the implementation of ASN.1 in Rosette were identified and corrected.
The Service Interface Agent abstraction was extended to support Service Agents
as responders as well as invokers. Planning is underway to deal with the
problems of heterogeneous configuration management of the diverse software that
makes up Carnot and must run on a variety of platforms. This will include
providing a direct Presentation Layer Interface in addition to the current
interface to the Remote Operation Service.
Interoperability via RDA
------------------------
Using Rosette as both a contract invoker and responder, a preliminary
implementation and testing of the following RDA operations has been completed:
rBeginDialogue, rEndDialogue, rBeginTransaction, rCommit, rRollback, rCancel,
rStatus, rOpen, rClose. In the responder side, no database is accessed, so the
Service Agent builds its own result structure based on the input argument
received. Also, the SQL Access Group tests are being looked at, to implement
requirements.
Graphical User Interface via HITS
---------------------------------
Now that CLM is working with Itasca, the next step is to see how to incorporate
this with other systems within Carnot. If it is feasible to incorporate
similar functionality into Rosette, Rosette could be enhanced to include the
interface elements into it - like the Motif library, allowing the Carnot
system to use Rosette as its core substrate for all types of interactions.
There are still some low level protocol issues to be ironed out to see if
this will be feasible.
Semantics-Based Graphical Languages
-----------------------------------
The group started to rework the underlying graphic representation for data
storage in the math scratchpad. This effort will be based on a quad-tree
representation for storing graphic objects in a 2-D plane. Math symbols
recognized or selected from a palette will be stored in this representation
and then various queries will access this data (locating nearest neighbors).
Hopefully, this will significantly speed up processing time.
They have incorporated the notion of residuation (from DEC engineers Ait-Kaci
and Nasr) into a revised design for unification in the graphical grammar
formalism.
Semantic Integration via CYC
----------------------------
Prototypes of two of the subsystems needed for semantic integration of
databases have been built: - Subsystem 1 parses a schema description, written
in Orion/Itasca syntax, for an object-oriented database and generates CYCL
commands for representing the schema in a CYC microtheory; - Subsystem 2
applies articulation axioms to translate a CYCL query, expressed in terms of
CYC concepts, into one or more subqueries, expressed in terms of local database
concepts.
Two technical reports have been released which deal with semantic integration
and the CYC knowledge base:
ACT-OODS-073-91 (Q) "Semantic Integrity and Integration Support in Carnot" by
Collet & Huhns. This paper presents an analysis of the problem of integrating
separately developed information resources. It focuses on the properties
used to represent the semantics of a resource and describes the approach
used in the Carnot project.
ACT-OODS-080-91 "Discovering Regularities from Large Knowledge Bases" by
W. Shen. This paper addresses the problem of discovering regularities in
large knowledge bases that contain many assertions in different domains.
Application of this work to the CYC knowledge base has shown some interesting
results.
Deductive Computing
===================
LDL
---
The new documentation for LDL has been completed. A reference number and title
will be provided in this space next month.
Work continues on the design document for LDL++. A draft release should be
available within the next one-two months and will be announced in this space
when it is available.
Formal Methods Transition Study
-------------------------------
The first set of reports surveying the current state of formal methods research
and applications has been released by the Formal Methods Transition Study
group. These are the available reports:
STP-FT-1 Overview and Status: Topics and Deliverables
STP-FT-2 State Transition Systems
STP-FT-3 Z: Notation and Use
STP-FT-4 Object-Oriented Programming and Formal Methods
STP-FT-5 Tools Overview and Taxonomy
STP-FT-6 A Comparison of Theorem Proving Environments
STP-FT-7 Techniques for Assuring Correctness of Code
STP-FT-8 The Role of Testing in Formal Methods
These reports are available, as are all other MCC reports, from RDVAX::MCC upon
request. It is suggested you read STP-FT-1 first to see the structure of the
material and the method used to expand it into separate reports.
The Formal Methods group welcomes your comments; specifically, they would like
feedback on the following topics:
o Which reports are of most interest to you? Where would you like more
depth/breadth?
o Identify process descriptions that could be used to show how formal methods
would change the process. Your favorite process descriptions are welcomed
as input.
o If there is major confusion about a particular topic, please contact the
principal (first) author.
o Provide additional pointers to material and experience which you think might
be useful to integrate into future reports.
You may transmit comments to the Formal Methods people either through this
author at AUSTIN::CHERUBINI or to the Formal Methods Study Manager, Susan
Gerhart, at [email protected].
-----------------------------------------------------------------------------
OVERVIEW
If you have suggestions for getting MCC technology into Digital or want to be
added to or removed from the distribution list, send mail to AUSTIN::KUHLMAN.
MCC technical reports are available through the Digital Library System or by
sending mail to RDVAX::MCC and including the TR number and name, your name,
badge number, cost center and mail-stop. Newsletters are also available on
three other MCC activities: Neural Networks, Artificial Intelligence and Optics.
Subscribe by sending mail to AUSTIN::KUHLMAN. Monthly reports from the MCC
Experimental Systems (ES Kit) project are posted in the VAX Notes conference
RDVAX::MCC-ES-KIT.
MCC is a cooperative enterprise whose mission is to strengthen and sustain U.S.
and Canadian competitiveness in information technologies. The objective is
excellence in meeting broad industry needs through application-driven research,
development and timely deployment of innovative technology. The ACT (Advanced
Computing Technology) program mission is to anticipate and fulfill requirements
for computer systems technology that is five to ten years in advance of the
state-of-the-art, while maintaining a balanced R&D portfolio that provides
participants with continuing technology leadership. The theme of user-friendly,
distributed, heterogeneous systems is the basis for ACT's next five year vision.
Distribution:
TO: Pat Roach@VBE
TO: Susan Sugar@MWO
TO: Steve Becker@AQO
TO: Ed Hurry@DVO
TO: STEVE DONOVAN@DLO
TO: DENNIS DICKERSON@DLO
TO: Czarena Siebert@HSO
TO: Mike Sievers@HSO
TO: Mike Willis@HSO
TO: Sherry Williams@HSO
TO: Dale Stout@HSO
TO: Tommy Gaut@HSO
TO: Tom Wilson@HST
TO: jim rather@HSO
|
| 310.2 | FWD: MCC AI Update | ULYSSE::ROACH | TANSTAAFL ! | Tue Apr 02 1991 13:12 | 183 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 01-Apr-1991 09:23pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD03@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC AI Update
MCC AI UPDATE No. 14 INTERNAL USE ONLY
********************
March 28, 1991 (150 lines)
Prepared by: Bill Kuhlman, MCC liaison, Austin, Texas
Introduction
============
AI Update attempts to inform the Digital community of activities and events
of interest in the MCC Artificial Intelligence Laboratory. Please feel free
to forward it within Digital. Additions and deletions to the distribution
may be sent to AUSTIN::KUHLMAN. To obtain copies of MCC technical reports
mentioned in the Update, send mail to RDVAX::MCC (include mail stop).
KBNL (Knowledge-Based Natural Language) and Machine Translation
===============================================================
KBNL In-Depth
-------------
The KBNL In-Depth technical meeting has been scheduled for July 23-24, 1991
at MCC in Austin, TX. The new generic, portable KBNL system will be described.
Anyone interested in possible applications of the systems is encouraged to
attend.
Natural Language
----------------
One problem with natural language systems is that it will probably never be
possible to build ones that are so complete that they can simply be used as
black boxes. Instead, it must be possible for system developers to tailor an
NL system to a particular application. So the ease with which someone can get
inside the system and make changes to it is a very important factor in the
usefulness of any NL system. Over the last several months, the KBNL group has
been working on a major revision of the syntactic component of KBNL to make
the formalisms for the description of grammatical information easier to use.
Completely new formal languages have been designed and are being implemented.
Two new languages will replace three older languages that are now in use, and
this will simplify the task of the linguist working with the KBNL system in
several ways.
First, the new languages are designed to be easier to edit in a lisp-friendly
text editor. Second, the new languages are clearer and easier to work with:
they were designed for readability as well as easy computational manipulation.
Third, the reduction of the number of languages considerably shortens the
linguist's learning curve. As an added benefit, the new languages are simpler
to parse than the older languages. The older languages were completely
general, very powerful, and computationally elegant, while the new languages
sacrifice some generality for user-friendliness. The new languages more
closely reflect the computational mechanisms used in parsing, they allow the
use of a new, potentially interesting type of unification, and they should be
easier to modify in support of the linguist/user. The new languages are being
installed in parallel with the old language in the coming month, and they
should replace the older languages shortly thereafter.
Machine Translation
-------------------
Another major effort this month was the design of an algorithm for machine
translation that handles cases in which the literal semantics of the source
text and the NATURAL translation into the target language don't match up. An
algorithm was defined that solves this problem for single lexical items
(e.g., translating English "know" into Spanish "conocer" or "saber"). One very
important property of this algorithm is that it has very well-defined hooks
into the domain KB that supports it. If the KB is weak, the algorithm will have
to make arbitrary choices in many cases. But as the KB grows so that it can
answer questions that require reasoning based on the discourse context, the
performance of the algorithm improves dramatically.
Text Retrieval
--------------
A third major thrust this month was on the interface to the Scan text retrieval
system. The new system makes it much easier for users to see what the system
is doing and to exploit the browsing capabilities that it offers during query
reformulation. The new system served as the basis for a new videotape that
shows both the monolingual version of Scan and a bilingual version in which
texts are retrieved both in the language of the original query as well as the
other language that the system knows.
NL & MT Tech Reports
--------------------
ACT-NL-068-91 "Reversible Machine Translation: What to Do When the Languages
Don't Line Up" by Barnett et al. This paper deals with issues that face
an interlingua-based, reversible machine translation system when the literal
meaning of the source text is not identical to the literal meaning of the
natural target translation. An algorithm for lexical choice is presented.
ACT-NL-081-91 "Discourse Pegs and the Representation of One-Anaphora" by
S. LuperFoy. Following Landman's pegs model of NP semantics, this paper
posits two parameters for classifying all anaphoric phenomena.
CYC (Large Scale Common Sense Knowledge Base
===
Work continued with Phill Apley (CRL) and Bob Kerns on creation of the CYC
interface to the DECStation 3100 under CLIM. A first pass version of the MUE
(graphical editor) under CLIM was written.
The main efforts on the knowledge entering front this month were concerned with
human ailments, containment, and weather. Time was also spent mapping over
various slots concerned with human occupied spaces (e.g., buildings) such as
those involving physical attributes and spatial relations, in the Building
Microtheory.
A utility was written that collects expressions (such as inheritance and
type-level information) into files of UNASSERT and ASSERT (Functional Interface
expressions), which can be edited and loaded. Among other things, this makes
it possible to safely (and fairly easily) move inheritance expressions from one
Microtheory to another. This month this was mainly used to move inheritances
dealing with physical attributes from the BaseKB to the NaivePhysics and
Building Microtheories.
Work is underway on a facility for "sweeping" the knowledge base. The Sweeping
interface allows the user to specify a predicate to be added to units, then uses
a set of heuristics to find units on which one would expect to see an assertion
using the predicate. The user can then add the suggested assertion. Some code
was written to simplify disjunctions statements, as a prelude to writing code
for determining when an agent can and cannot do two things at the same time.
The technical report "The World According To CYC, Part 4" (ACT-CYC-022-91),
released last month, contains a section on the efforts in this area.
CYC-Related Technical Reports
-----------------------------
ACT-CYC-085-91 "Index to CYC Project Publications (Version 1.1)" by Shepherd.
This index is intended to facilitate the use of reports from the CYC project.
It includes a guide to using the index, a listing of the items indexed, a
copy of the abstract for each technical report, and the preface to the book
"Building Large Knowledge-Based Systems".
ACT-OODS-080-91 "Discovering Regularities from Large Knowledge Bases" by
W. Shen. This paper addresses the problem of discovering regularities in
large knowledge bases that contain many assertions in different domains.
Application of this work to the CYC knowledge base has shown some interesting
results.
================================================================================
Newsletters are also available on three other research areas at MCC:
Systems Technology, Neural Networks, and Optics. To subscribe, send mail to
AUSTIN::KUHLMAN. Reports from the MCC Experimental Systems project ES-Kit are
posted in the VAX Notes conference RDVAX::MCC-ES-KIT. For reports on MCC's
work on Packaging/Interconnect and High Temperature Superconductivity, contact
Chauvet Parker (AUSTIN::PARKER).
MCC is a cooperative enterprise whose mission is to strengthen and sustain U.S
and Canadian competitiveness in information technologies. The objective is
excellence in meeting broad industry needs through application-driven research,
development and timely deployment of innovative technology. The ACT (Advanced
Computing Technology) program mission is to anticipate and fulfill requirements
for computer systems technology that is five to ten years in advance of the
state-of-the-art, while maintaining a balanced R&D portfolio that provides
participants with continuing technology leadership. The theme of user-friendly,
distributed, heterogeneous systems is the basis for ACT's next five year vision.
Distribution:
TO: Pat Roach@VBE
TO: Susan Sugar@MWO
TO: Steve Becker@AQO
TO: Ed Hurry@DVO
TO: STEVE DONOVAN@DLO
TO: DENNIS DICKERSON@DLO
TO: Czarena Siebert@HSO
TO: Mike Sievers@HSO
TO: Mike Willis@HSO
TO: Sherry Williams@HSO
TO: Dale Stout@HSO
TO: Tommy Gaut@HSO
TO: Tom Wilson@HST
TO: jim rather@HSO
|
| 310.3 | FWD: ANNOUNCEMENT - Smith on ESP, 4/5, 10am | ULYSSE::ROACH | TANSTAAFL ! | Wed Apr 03 1991 18:16 | 94 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 02-Apr-1991 06:58pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD05@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: ANNOUNCEMENT - Smith on ESP, 4/5, 10am
From: AUSTIN::KUHLMAN "Bill Kuhlman (512) 338-3243 02-Apr-1991 0018" 2-APR-1991
00:21:44.11
To: @[KUHLMAN.SYSTEMS]SYSNEWS
CC: KUHLMAN
Subj: ANNOUNCEMENT - Smith on ESP, 4/5, 10am
TITLE: EXTENSIBLE SOFTWARE PLATFORM
SPEAKER: Kim Smith
Microelectronics and Computer Technology Corporation (MCC)
DATE: Friday, April 5, 1991
TIME: 10:00 - 12:00 A.M.
PLACE: Babbage Auditorium, ZKO1-3, Spitbrook Rd., Nashua, N.H.
HOST: Liz Freburger, Object-Oriented Program, TNSG
This presentation will discuss the C++ object-oriented paradigm as the
enabling technology for ESP. The current state of the ESP implementation
will be desribed and the direction and objectives of current and future
research plans will be discussed. The significant technical points to be
covered include:
1 - Use of the apparent semantics of standard C++ to allow concurrency,
2 - Use of a "FUTURES" class to provide for synchronization and exception
handling,
3 - Use of the "placement argument" for NEW to allow expression of an
application dependent Object Placement strategy in a distributed
or parallel system at compile time.
4 - Development of automatic placement algorithms to provide tactical
placement implementation at execution time.
5 - Application experiences with the ESP software, and
6 - Plans for commercialization through MCC participants.
The three year, $6M DARPA-funded Experimental Systems Project (ES-Kit)
nears the end of its third year. Besides the development of a small
set of MC88000-based hardware building blocks, ES-Kit technology
includes a maturing micro-kernel that supports the execution of
object-oriented applications on a distributed, heterogeneous platform.
The ES-Kit micro-kernel is currently being used in a number of
University-centered research efforts and commerical adaptations.
At the end of this year, ES-Kit will evolve into two separate efforts.
One of these, Advanced High Performance Modules (AHPM), is a DARPA-
funded continuation of the development of high performance mesh based
parallel computing hardware(100MIPS, 64MB and 20MB/sec message mesh in
a business-card sized package). The other effort, the Extensible Soft-
ware Platform (ESP) will continue development of the ES-Kit software on
behalf of commercial and academic sponsors, currently being solicited.
The ESP project is being defined as a three pronged effort including rapid
maturation of the ES-Kit baseline technologies, significant application
studies, and technology transfer.
-------------------------------------------------------------------------
K. Stuart Smith joined MCC's Parallel Processing Program in 1986 following
seven years working in the supercomputing arena as a system programmer and
computer architect. Since joining Experimental Systems in 1987, Mr. Smith has
been responsible for the advanced design and implementation of the ES-Kit
"kernel" and system software, where current emphasis is placed on generalized
functionality and improved reliability. Prior to 1986, Mr. Smith was employed
by Burroughs Corporation and American Supercomputers. He has taught graduate
level courses in operating systems and computer architecture at Villanova
University. Mr. Smith holds the M.S. degree in computer science from the
University of Colorado (1979) and a B.A. in chemistry from the State
University of New York (1976).
Distribution:
TO:
Steve Becker@AQO DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO
Ed Hurry@DVO jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.4 | FWD: March STP 1991 MCC Technical Report Abstracts | ULYSSE::ROACH | TANSTAAFL ! | Tue Apr 09 1991 15:11 | 54 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 08-Apr-1991 06:00pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD05@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: March STP 1991 MCC Technical Report Abstracts & Revisions
For Digital Internal Use Only
March, 1991 MCC Technical Report Abstracts & Revisions
Title: The LDL Query Optimizer: Architecture
Author(s): Brijesh Agarwal, Ravi Krishnamurthy, Arshad Martin,
Tom McLellan, Leona Slepetis, Peter Song
Number: STP-LD-001-91-(P)
Date: March, 1991
This document describes the architecture of the LDL Optimizer, an important
module of the system which supports the compilation and execution of
applications written in the logic-based, non-procedural language LDL. The
function of the Optimizer is to devise a safe and efficient execution plan for
a given LDL program and to convey that plan to the Translator. The input LDL
program is represented as a relevant PCG, and the execution plan is conveyed
to the Translator in the form of a controlled PCG. The Optimizer makes a
number of decisions that lead to the selection of an optimal execution plan
for a given query form. These decisions concern: (A) the ordering of goals in
the body of a rule (join order), (B) the join method to be used for each of
the joins, and (C) the method to be used for compiling each recursive goal.
Join method choices include: (1) whether to materialize or pipeline
intermediate relations, (2) whether or not to reselect, (3) whether or not to
use any existing indices, and (4) whether or not to create an index.
Compilation method choices for recursive goals include semi-naive, magic, and
counting methods. This paper describes the techniques used in the
implementation, along with the formulas used in estimating the cost of
queries.
================================================================================
Distribution:
TO:
Steve Becker@AQO DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO
Ed Hurry@DVO jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.5 | FWD: MCC AI Update | ULYSSE::ROACH | TANSTAAFL ! | Mon May 06 1991 11:08 | 161 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 01-May-1991 06:56pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD20@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC AI Update
MCC AI UPDATE No. 15 INTERNAL USE ONLY
********************
April 24, 1991 (137 lines)
Prepared by: Bill Kuhlman, MCC liaison, Austin, Texas
Introduction
============
AI Update attempts to inform the Digital community of activities and events
of interest in the MCC Artificial Intelligence Laboratory. Please feel free
to forward it within Digital. Additions and deletions to the distribution
may be sent to AUSTIN::KUHLMAN. To obtain copies of MCC technical reports
mentioned in the Update, send mail to RDVAX::MCC (include mail stop).
CYC (Large Scale Common Sense Knowledge Base)
===
A Project Review meeting of the CYC project will be held at MCC on May 7, 1991
8:30 am - 1:00 pm. The Review is open to all interested project participants.
Digital personnel are invited. Contact Bill Kuhlman if you plan to attend.
The main efforts on the knowledge entering front this month were concerned with
preparing vocabulary for the weather domain as well as entering weather
knowledge into the knowledge base. For example, information was added about
clouds, seasons, annual climate cycles, storms, etc. In preparation for the
Automobile Advisor Demo, which will be ready by the May 7 project review,
knowledge was entered in AutoMt (the Microtheory dealing with automobiles)
pertaining to how geographic location, kinship, and hobbies might affect a
person's car-buying decisions.
Work began on a naive theory of manufacturing, as well as a shell for providing
a natural language interface to applications built in CYC.
There were five papers presented by members of the CYC Project at the AAAI
Spring Symposium at Stanford. Four of them were given in the session on
Logical Formalisms for Common Sense. ("Naive Theory of Money", by Dexter Pratt
and Wanda Pratt, "Information Bearing Things," by Guha and Pittman,
"Counterfactuals" by Guha and Lenat and "Performing Concurrent Actions" by
Goolsbey and Guha). The other was presented in the Implemented Knowledge
Representation Systems session ("Evolution of CYCL", by Guha and Lenat).
KBNL (Knowledge-Based Natural Language) and Machine Translation
===============================================================
The KBNL knowledge-based natural language processing system is novel in the way
it is connected to backend knowledge bases via a generic NL/KB interface, in
the clean separation it enforces between linguistic knowledge and world
knowledge, in its use of knowledge to aid in lexical acquisition, and in its
exploitation of a single knowledge base to support both language understanding
and language generation. The following MCC technical report describes the KBNL
system and three applications - text retrieval, machine translation, and the
combination of them--multilingual text retrieval:
ACT-NL-123-91 "Knowledge-Based Natural Language Processing: the
KBNL System" by Barnett et al.
One of the major technical goals of the work on Lucy and Koko has been
reversibility. If ways can be found to represent linguistic knowledge once and
use it in one direction for understanding in Lucy and in the other direction
for generation in Koko, then the cost of both building and maintaining those
systems is reduced, as compared with more conventional approaches that require
separate specifications for each direction.
Software release
----------------
A new version of the KBNL systems is being released. This is the first release
of a portable KBNL. The parts of the system that depend on platform-specific
capabilities (for example, multiprocessing) have been parameterized and all
the platform-specific code has been collected into one place. The released
system has been tested on three platforms: Symbolics, Lucid Common Lisp (on
Sun 3's and Sparc Stations) and Allegro Common Lisp (on Sparc Stations).
KBNL Port planned to DECStation
-------------------------------
Digital has provided a DECStation 3100 to the KBNL group. Work on a port of
the new system will begin next month.
In order to make the new system usable on a variety of platforms, the user
interface was reimplemented, both to the NL developer's toolkit and to Scan.
The new interfaces to both systems have been built using Garnet, a toolkit
developed at CMU. Garnet compiles into either X Windows or Macintosh windows,
so the interface will be able to run on any platform that supports one of those
systems. The Garnet interface that is being released has been designed to
provide basic functionality. It does not, however, provide all of the user
options that were available in the Symbolics version. The KBNL team expects
the interface to grow over the next several months, and they are particularly
interested in input from users to help determine what kinds of capabilities
would be most useful.
KBNL meetings
-------------
On July 22 and 23, 1991 the KBNL project will have an indepth meeting for
MCC shareholder participant personnel to talk about the new system and its
possible applications. An updated version of the system running on the DS3100
will be available at that time.
In addition to this indepth technical meeting, a strategic planning meeting
is scheduled for May 8. This meeting will bring together representatives
from the current KBNL participants, other MCC shareholders who have expressed
a strategic interest in natural language, and other organizations (including
government ones) that have expressed an interest in participating in this
effort. The goal of this meeting is to explore the range of application areas
for NL (including text retrieval, interfaces to databases, machine translation,
backends to speech systems, etc.) and to decide where, within this space,
efforts should be focused in order to have the most impact on project
participants.
Natural Language Technical Reports
----------------------------------
ACT-NL-067-91 "Shared Preferences" by Barnett & Mani. This paper attempts to
develop a theory of heuristics or preferences that can be shared between
NL understanding and NL generation systems.
ACT-NL-117-91 "Resolution of Collective-Distributive Ambiguity Using Model -
Based Reasoning" by C. Aone. This paper presents a semantic analysis of
collective-distributive ambiguity, and the resolution of such ambiguity by
model-based reasoning.
================================================================================
Newsletters are also available on three other research areas at MCC:
Systems Technology, Neural Networks, and Optics. To subscribe, send mail to
AUSTIN::KUHLMAN. Reports from the MCC Experimental Systems project ES-Kit are
posted in the VAX Notes conference RDVAX::MCC-ES-KIT. For reports on MCC's
work on Packaging/Interconnect and High Temperature Superconductivity, contact
Chauvet Parker (AUSTIN::PARKER).
MCC is a cooperative enterprise whose mission is to strengthen and sustain U.S
and Canadian competitiveness in information technologies. The objective is
excellence in meeting broad industry needs through application-driven research,
development and timely deployment of innovative technology. The ACT (Advanced
Computing Technology) program mission is to anticipate and fulfill requirements
for computer systems technology that is five to ten years in advance of the
state-of-the-art, while maintaining a balanced R&D portfolio that provides
participants with continuing technology leadership. The theme of user-friendly,
distributed, heterogeneous systems is the basis for ACT's next five year vision.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.6 | FWD: MCC Systems Report | ULYSSE::ROACH | TANSTAAFL ! | Mon May 06 1991 11:10 | 649 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 01-May-1991 06:56pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD05@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC Systems Report
MCC Systems Technology Report INTERNAL USE ONLY
*****************************
April 25, 1991 (624 lines. Last
500="FM Today")
Prepared by Bill Kuhlman and Ralph Cherubini, Austin, TX
This newsletter covers the work in the Object-Oriented and Distributed Systems
(OODS) lab within MCC's Advanced Computing Technology (ACT) Program and the
Deductive Computing Lab within the Software Technology Program (STP). Please
feel free to forward it within Digital. Address inquiries concerning Deductive
Computing projects to AUSTIN::CHERUBINI; OODS projects to AUSTIN::KUHLMAN.
OODS (Carnot)
=============
Review Meeting - May 7
----------------------
A Project Review meeting of the Carnot project will be held at MCC on
May 7, 1991, 1:00-5:30. The meeting is open to Digital personnel interested
in evaluating the project for potential support. Contact Bill Kuhlman if
you plan to attend.
Video Tapes available
---------------------
A marketing video of "The Carnot Project" is available as an MCC technical
report (ACT-OODS-101-91). It is a good, short (13.5 min) summary of the
goals and directions of the project.
The video (40 min) of the technical seminar which Phil Cannata presented at ZKO
on March 13 is available from the ZK library (WECARE::TNSG_TRAIN).
Carnot Technical Progress
-------------------------
Distributed Transaction Environment via Rosette: Implemented the SQL Access
Group's version of the Remote Data Access protocol in Rosette; Implemented a
Service Interface Agent and Service agent that manages the Itasca Database;
Installed ISODE 6.8 and integrated Rosette with it; Will be upgrading to
Sunlink OSI 7.0 in anticipation of participating in the SAG interoperability
demonstration. Rosette has been modified to support the ISO8859 character set
as required by the SQL Access Group.
SAG Interoperability via RDA: Prepared Rosette scripts for tests A and B of
the SQL Access Group.
SAG OODB Server: Constructed the "intergalactica" database in Itasca; Began
implementation of SQL query front end for Itasca; Implemented a Rosette RDA
Service Agent for Itasca.
SAG Client Applications: The Carnot Graphical Interaction Environment was
connected to Rosette. This combination will serve as the SAG client application
environment.
Preparations for the July 1991 SAG Interoperability Demo: A stand-alone
environment is being set up and will be connected to the interoperability lab
at HP in California via a lan-bridge. The next several weeks will be
spent conducting interoperability tests with the other demo participants.
The bridge will also be used to conduct interoperability tests with
participants that are not at the HP lab, such as NCR in San Diego.
Semantics-Based Graphical Languages: extended the parsing algorithm to handle
graphical grammars where relations hold between parts of objects
(instead of just whole objects).
Semantic Integration via CYC: built prototypes of two more of the subsystems
needed for semantic integration of databases. Subsystem 1 parses simple
database queries written in a subset of SQL and translates them into CYCL.
Subsystem 2 parses subqueries written in CYCL (as produced by the subsystem
that applies articulation axioms) and translates them into SQL for execution on
a database.
Experimental Systems
====================
The Experimental Systems lab received notification from DARPA that the Advanced
High Performance Modules project will be funded (approx. $9 million). This
is the follow on to the ES-Kit parallel processing project.
Comings and Goings
===================
Phil Cannata spoke on the Carnot project at Digital's Japan R&D Center,
Yokohama, on April 3.
Kim Smith presented a technical seminar on MCC's Extensible Software Platform
(ESP) at ZKO on April 5.
Damir Jamsek, from MCC's Formal Methods Transition Study visited the C++
group in ZKO on April 15.
Deductive Computing
===================
LDL
---
The new LDL User's Guide will be completed THIS month. I will send out a
technical report number as soon as it becomes available.
Shalom Tsur is drafting an NSF proposal to create a knowledge base for the
bacterium, E. Coli, which would include representations of biochemical
reactions as well as genetic sequences.
LDL is now supported by a Sybase interface, complementing the Rdb interface
already available.
A MCC technical report on LDL is available: STP-LD-001-91(P) "The LDL
Query Optimizer: Architecture" by Brij Agarwal.
Formal Methods Transition Study
===============================
The current issue of FM Today is reproduced below:
+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-
+ +
+ FM Today +
+ ======== +
+ Newsletter of the MCC Formal Methods Transition Study +
+ March 1991 +
+ +
+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-+*-
Contents
--------
The VIPER Controversy in the UK
1991 Conferences (previous + ICSE + Compass'91)
FMTS Report Titles
The VIPER Controversy in the UK
-------------------------------
The VIPER (Verifiable Integrated Processor for Enhanced Re-
liability) Microprocessor appeared on the scene in 1988 with
considerable fanfare as one of the first major applications
of formal verification to realistic problems. During 1990,
several problems with VIPER's claims became widely known,
leading to a serious dispute among its various research and
development parties. The purpose of this FM Today article is
to first apprise you of the situation as we see it and then
identify issues that we consider important for the future of
North American formal methods.
First, a recap. The Royal Signal and Radar Establishment
(RSRE) in Malvern England has been a research center for
over a decade in formal methods (broadly construed as
specification, verification, and semantic analysis). During
the mid-1980s, the British Ministry of Defence became con-
cerned that microprocessors they were procuring and using in
defense applications did not match their manuals and could
not be trusted in action. Therefore, a more precise specifi-
cation and verification process became highly attractive. At
the same time, the MoD was engaged in an effort to commer-
cialize many of its research results through the Alvey and
other initiatives. This spawned smaller organizations
licensing the technology and centers of users in larger com-
panies. Charter Technologies obtained the license to market
VIPER and develop further tools for it.
Conferences and review meetings were held at RSRE and in
London and a newsletter, SafetyNet, was published. SafetyNet
contained tutorial articles on formal methods and VIPER by
R&D personnel in companies such as Praxis, Rolls Royce,
Plessy, Program Validation Ltd., Marconi, Rex Thompson
Partners, and British Aerospace.
---------------------------------------------------------
---------------------------------------------------------
VIPER TOOLS
-----------
VISTA: High Level Assembly Language to VIPER
VISOR: Simulator of VIPER
MALPAS and SPADE: static analyzers
STEVE: VISTA-to-SPADE translator
MAVIS: VISTA-to-MALPAS Intermediate Language translator
SPADE-PASCAL: Pascal subset for safety-critical applications
VIPSE: Integrated Programming Support Environment
Trademarks: VIPER U.K. MoD; SPADE, Program Validation Limited
---------------------------------------------------------
---------------------------------------------------------
A Peer Review was undertaken, mainly focused on the limita-
tions of the processor - functionality, shortage of regis-
ters, word length, interrupts, performance - noted as re-
quirements for future versions of the chip. There does not
appear to have been any comparable review of the verifica-
tion claims, although use of VIPER by the Peer Reviewers
didn't report any inconsistencies. In addition to Charter
running the educational and PR symposia and events, Marconi
Electronic Devices received the contract for a commercial
prototype. Charter began developing systems using the pro-
cessor for a non-safety critical system and a major railway
signaling system to be built by Teknis, an Australian com-
pany.
Meanwhile, back in the research labs at Cambridge Universi-
ty, the proof efforts were being wrapped up. The researcher,
Dr. Avra Cohn, having finished the contracted proofs, pub-
lished a full account in Cambridge Technical Reports and a
journal article. Two kinds of problems were uncovered or
more fully recognized during this proof process. First, the
technical aspects of the proof included only instruction
state transitions (fetch-decode-execute cycle but not reset
or timeout or arithmetic). Second, even had more of the
proof been done, there were limits on any claims that could
be made: neither an actual device nor the intentions for its
behavior could be subjected to logical reasoning, and there-
fore must be accepted as boundaries of a proof and subjected
to further scrutiny by other means. Furthermore, the
designers, verifiers, and manufacturers were neither cen-
trally managed nor working off the same documents or common
models. The mechanical proofs that were completed and their
real meaning are cogently summarized in Cohn's technical re-
ports, but certifying such proofs is ill-defined within the
research process.
The fully completed proof would have looked like:
hand proof:
top-level to major-state
Simple state-transition function representing visible
memory and register changes as instructions execute
in HOL:
major-state level relative to top-level
(equivalence of effects of sequences of events to
top-level instruction step)
Major state level with sequence of events rather than
single step, operating on an internal state (but still
abstract from actual chip)
in HOL:
block-level relative to major-state
(deriving mathematical function for block level,
proving the equivalence with major states)
Block model in text and pictures for the instruction
decoder, ALU, and memory on fixed clock cycles
ELLA simulation:
Block level to implementation
using "intelligent exhaustion"
Shortly thereafter, NASA Langley Research Center, having
similar interests in formally verified chips and consider-
able experience with managing fault-tolerance system proofs
(SIFT), commissioned a report by Computational Logic, Inc.,
in Austin, Texas. CLI researcher Warren Hunt had proved pro-
perties of a non-manufactured chip, the FM8501, as his doc-
toral dissertation at UT using the Boyer-Moore prover tech-
nology. The NASA report dissects the VIPER proof effort,
compares it with the UT and subsequent CLI work, and inter-
views the VIPER designers. The findings were critical of the
VIPER claim to the status of "fully verified microproces-
sor".
Charter Technologies learned of these results only indirect-
ly (not from MoD) in 1989. After claiming that RSRE would
not respond to their concerns, in October 1990 Charter
threatened suit and in January 1991 began legal proceedings
against the MoD. Articles appeared soon in the UK Sunday
Correspondent, the Independent, the Engineer, Computing, and
other trade publications and newspapers. Charter also
claimed that MoD had overstated the commercial potential of
the processor and the number of potential customers.
Here's our assessment of the situation. This is the first
major test in the courts of law (and reputation) of claims
of a "formally verified anything". However, we see the many
claims involved as blatantly false, although the overall ef-
fort is very well-intentioned with no evidence of deceit.
Some of these claims are:
1. A 1987 letter by RSRE Director N. H. Hughes, says
"VIPER is the first commercially available microprocessor
with a proven correct design... By combining VIPER with nov-
el means of designing and proving the correctness of comput-
er programs, also developed by RSRE and collaborating com-
panies, it will be possible to produce much safer micropro-
cessor based systems in the 1990s and beyond". RSRE's pride
in their microprocessor is clearly shown and justified. How-
ever, unless scrutinized by some kind of external peer re-
view process, including both hardware and verification ex-
perts, a lab director's claim of "proven correct" holds no
more meaning than that "RSRE says so". Furthermore, there
are many different meanings that can be attached to the term
"proven correct design", as in: "proofs were done", "proven
to hold up under use", "proven to design engineers' satis-
faction", etc. Maybe RSRE just didn't understand these
nuances of the term "proven".
2. Charter claims that RSRE did not reveal the Cam-
bridge U. caveats about the levels and degrees of proofs ac-
tually achieved. MoD rejects Charter's suit on the basis
that Charter should have been aware of the reports.
Charter's statements in the press hint at some suppression
of reports from Charter's attention. It seems incredible
that RSRE and Cohn did NOT forward the reports to Charter,
but we all know how bureaucracies work - "somebody else is
doing it" or "it just hasn't been done yet". It also seems
incredible that Charter, having undertaken business respon-
sibility for a very advanced technology did not perform "due
diligence" in checking in with the researchers about their
views. Also, Charter had considerable contact with the UK
R&D community at its various functions and through its
newsletters. It seems that Charter Technology did not "fol-
low the field".
3. The research reports from Cambridge University and
CLI provide full accounts of the claims from a verifier's
perspective. The CLI/NASA report must also be read as coming
from a competitor (CLI's chip had not yet been manufactured)
and Cohn as coming from a very conservative researcher. For
example, Cohn discusses the huge amount of effort involved
in one of the proofs - six months of grueling work. While
six months is a long time in the life of a researcher, it's
a drop in the bucket to the amount of testing and verifica-
tion effort commonly employed on modern hardware systems.
4. Some of the newspaper accounts are quite sensation-
alized "Nasa says `fault-free' MoD chip is useless" We
couldn't find any basis for that in the CLI report. As usu-
al, in highly technical material as involved here, newspaper
accounts will be only partially accurate.
Now, what does this all say for the future of formal
methods?
1. The VIPER dispute addresses only one small but sig-
nificant aspect: formal verification of low-level hardware
system functionality, basically state machine interpreters
of the microprocessor instruction set. Formal specification
for the purpose of system definition and validation are not
affected by these claims, nor are the proofs of many securi-
ty and other properties that are understood to be limited to
certain aspects of a system specification.
2. The VIPER dispute crystallizes many issues that have
long concerned researchers in the verification field - what
does a proof mean? how much of a proof must be accomplished
to supply some certainty against some kinds of failures? how
should that certainty be interpreted? "Verification" is a
meaningless stand-alone term - some property was proved re-
lating some pair of well-defined objects by some means
within certain limits of current knowledge and bounds of the
context. Long-standing confusion about these issues keeps
re-occurring, because the issues are not only philosophical,
but also engineering and scientific.
3. What would the VIPER proof have required to be suc-
cessful? Well, first it would have taken many years, given
the current state of mechanical provers, and by hand it
would have been on the level of difficulty of the early com-
putations carried out during World War II before computers.
Second, a more precise statement of what would be proved
should have been made, agreed upon independently of the de-
tails, and tracked throughout the proof effort. Third, the
proof would have been performed iteratively, and possibly
replicated on other provers (this is feasible and an in-
teresting research topic). What was accomplished at Cam-
bridge was basically just the first iteration.
4. The "stack", "tower", or "ladder" approach of
multi-level proofs is becoming a more focused research to-
pic. CLI makes it a part of their research and business
strategy and a major ESPRIT Basic Research Action (ProCoS)
is addressing alternative approaches. Again this addresses
only part of the FM research agenda, ignoring the assurance
techniques required at the top of the ladder, but it would
be a major step forward to have more verifiable and verified
system software.
On the personal side, we have considerable background in the
VIPER dispute. Ted Ralston and Susan Gerhart, on their MCC-
ILO sponsored UK tour in 1988, visited RSRE at the end of
one of the first VIPER symposia and sampled the many vendor
booths and a few talks. There was an air of both hype and
optimism - a new technology was clearly in the making. John
Cullyer, member of the VIPER design team and now a Professor
at Warwick University, visited MCC during his 1989 visit to
Austin for the CLI report and gave a quick tutorial on his
system engineering approach, including some interesting but
ill-formed ideas for hybrid verification and validation
techniques mixing testing and proving. Being planned is a
joint meeting of CLI, the ProCoS project (involving the ma-
jor FM luminaries Hoare and Bjorner), and the MCC formal
methods projects, mainly a multiple technical interchange.
And, we have also been in contact with an interesting UK
research project on the "sociology of mathematics", which is
looking at the court-room aspects of proofs and the negotia-
bility of mathematics as in the floating point chip stan-
dards, as well as issues of professionalization and the role
of formal methods in that movement.
So, the bottom line is that the VIPER dispute is clearly a
milestone in some respect, but in no way an industrial-
strength application of formal verification. We hope it will
clear the air about unwarranted claims for verifiability
while remaining in the perspective of the whole range of
techniques being explored and their current practicality.
Certainly, any FMTS participant expecting to make any future
claims should keep in mind the above lessons: get lots of
review from different quarters before making any claims,
stay in touch with the literature and the research communi-
ty, and recognize the current adage that "something too good
to be true probably isn't true".
REFERENCES: We have a fairly complete file of material at
MCC, including product literature, SafetyNet newsletter,
clippings from the January 1991 newspaper articles, and the
Cambridge University and NASA/CLI technical reports.
Avra Cohn
---------
The notion of proof in hardware verification. Journal of Au-
tomated Reasoning, 5, 1989.
A proof of correctness of the VIPER microprocessor: the
first level. In VLSI Specification, Verification and Syn-
thesis, Kluwer, 1987. (Also Cambridge U. Computer Laboratory
TR No. 104).
Correctness properties of the VIPER block model: the second
level. In Current Trends in Hardware Verification and Au-
tomat ed Deduction, Springer Verlag, 1988. (Also Cambridge
U. Computer Laboratory TR No. 134).
John Cullyer
------------
Implementing safety-critical systems: the VIPER microproces-
sor. In VLSI Specification, Verification and Synthesis,
Kluwer, 1987.
Warren A. Hunt, Jr.
------------------
FM8501: A Verified Microprocessor. TR 47, University of Tex-
as, 1985.
The Mechanical Verification of a Microprocessor Design. TR
CLI-6, Computational Logic, 1987.
(with Bishop Brock). Report on the Formal Specification and
Partial Verification of the VIPER Microprocessor. TR 46,
Computational Logic Inc., January 1990.
Safetynet -
---------
VIPER microprocessors in high integrity systems.
VIPER Technologies Ltd., Worcester England. (several issues
published in 1988 and 1989).
Donald MacKenzie
----------------
Formal Methods and the Sociology of Proof. In British Com-
puter Society Formal Aspects of Computer Science Refinement
Workshop, 9-11 January 1991, to appear in Springer Verlag
Lecture Notes in Computer Science.
The VIPER Microprocessor: A Study in the Sociology of Proof,
March 1991, preliminary report.
Conferences
-----------
For more details, see FM Today Issue No. 3
Symposium on Testing, Analysis and Verification, Victoria,
British Columbia, Canada, October 8-10, 1991. (ACM Sigsoft)
VDM'91, Formal Software Development Methods,
Noordwijkerhout, The Netherlands, October 21-25, 1991.
International Workshop on Software Specification and Design,
Lake Como, Italy, October 25-26, 1991. (IEEE Computer So-
ciety and ACM Sigsoft).
13th International Conference on Software Engineering, Aus-
tin Texas, May 12-17, 1991:
*******Formal Methods in Software Engineering, May 12. Mark A.
Ardis, Carnegie Mellon University Software Engineering In-
stitute, and Susan L. Gerhart, MCC Software Technology Pro-
gram.
Survey of Formal Methods: Historical perspective, Applica-
tion experience. Method Sampler: Foundations, technology,
lifecycle considerations, comparisons and strengths and
weaknesses (Larch, VDM, Z, state-machine methods). Survey of
tool support for formal methods: taxonomy, uses, demonstra-
tion, comparisons and strengths and weaknesses. Case Studies
of applications: industrial uses, project characteristics
and data.
Registration Information: Barbara Smith, 512-338-3336, FAX
512-338-3899, email [email protected].
*******MCC Formal Methods Open House. May 14, 4-6 p.m. Discussions,
Demos, Refreshments.
COMPASS `91 6th Annual Conference on Computer Assurance, Na-
tional Institute of Standards and Technology, Gaithersburg,
MD, June 24-28, 1991:
Featured Speakers: David L. Parnas, Queens University; John
Cullyer, University of Warwick; H. O. Lubbes, Naval Research
Laboratory.
Tutorials: "Safe Systems: A Disciplined Approach" (McDermid,
University of York and Cullyer, University of Warwick);
"Software Safety Analysis: Linking Fault Trees and Petri
Nets" (Gill, Naval Air Test Center).
Panel: "Educating Computer Scientists for the Year 2000":
Cherniavsky, NSF; Parnas, Queens U.; Denning, NASA Ames
Research Center; Scherlis, DARPA; McDermid, University of
York/British Computer Society.
Talks: "The Accidents of Life-From Conception to Our Last
Moments", John Cullyer, University of Warwick; "Computer Re-
lated Risk of the Year: Weak Links and Correlated Events",
Peter Neumann, SRI.
Birds of a Feather: "Software Development Methods in Prac-
tice", J. V. Hill, Rolls-Royce and Associates Limited
Keynote: "High Assurance Computing", H. O. Lubbes, Naval
Research Laboratory
Sessions (Selected Papers):
EUROPEAN ECONOMIC COMMUNITY `92 PERSPECTIVES ("Airbus A320
Software Safety", "Some Results From DRIVE")
HOW INDUSTRY TRAINING IN COMPUTER ASSURANCE CAN BE IMPROVED
THROUGH EDUCATION
CERTIFICATION AND SAFETY OF CRITICAL SYSTEMS
FORMAL METHODS ("Report on the Formal Specification and Par-
tial Verification of the VIPER Microprocessor" (Brock and
Hunt, CLI); "Using Correctness Results to Verify Behavioral
Properties of Microprocessors" (Windley, Idaho); "Estella: A
Facility for Specifying Behavioral Constraint Assertions in
Real-Time Rule-Based Systems" (Browne and Mok, UT); "Design
Strategy for a Formally Verified Reliable Computing Plat-
form" (Butler and Caldwell, NASA Langley Research Center);
"Developing Implementations of Estelle Specifications Using
the PEDS Toolkit" (Majurski, NIST).
US AND INTERNATIONAL SPONSORED INITIATIVES: "NIST: Workshop
on Assurance of High Integrity Software"; "NASA Langley:
Research Program in Formal Methods".
RISK CONTAINMENT PLANNING AND QUALITY MEASUREMENTS
Forum: US and International Standards for High Integrity
Systems (DoD, Government, and Industry)
***********************************************************
* Tentative dates -- next FMTS review meeting *
* June 18-20 *
***********************************************************
FMTS Reports:
------------
Participants should contact their FMTS technical representa-
tives for copies of the following: (Digital people should
contact RDVAX::MCC to request copies).
STP-FT-1 Overview and Status: Topics and Deliverables
STP-FT-2 State Transition Systems
STP-FT-3 Z: Notation & Use
STP-FT-4 Object-Oriented Programming and Formal Methods
STP-FT-5 Tools Overview and Taxonomy
STP-FT-6 A Comparison of Theorem Proving Environments
STP-FT-7 Techniques for Assuring Correctness of Code
STP-FT-8 The Role of Testing in Formal Methods
-----------------------------------------------------------------------------
OVERVIEW
If you have suggestions for getting MCC technology into Digital or want to be
added to or removed from the distribution list, send mail to AUSTIN::KUHLMAN.
MCC technical reports are available through the Digital Library System or by
sending mail to RDVAX::MCC and including the TR number and name, your name,
badge number, cost center and mail-stop. Newsletters are also available on
three other MCC activities: Neural Networks, Artificial Intelligence and Optics.
Subscribe by sending mail to AUSTIN::KUHLMAN. Monthly reports from the MCC
Experimental Systems (ES Kit) project are posted in the VAX Notes conference
RDVAX::MCC-ES-KIT.
MCC is a cooperative enterprise whose mission is to strengthen and sustain U.S.
and Canadian competitiveness in information technologies. The objective is
excellence in meeting broad industry needs through application-driven research,
development and timely deployment of innovative technology. The ACT (Advanced
Computing Technology) program mission is to anticipate and fulfill requirements
for computer systems technology that is five to ten years in advance of the
state-of-the-art, while maintaining a balanced R&D portfolio that provides
participants with continuing technology leadership.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.7 | Fwd: April 1991 MCC Technical Report Abstracts & Re | ULYSSE::ROACH | TANSTAAFL ! | Mon May 06 1991 11:16 | 269 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 03-May-1991 10:48pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD20@DPD
Dept:
Tel No:
TO: See Below
Subject: Fwd: April 1991 MCC Technical Report Abstracts & Revisions
For Digital Internal Use Only
April, 1991 MCC Technical Report Abstracts & Revisions
Title: Shared Preferences
Author(s): Jim Barnett, Inderjeet Mani
Number: ACT-NL-067-91
Date: February, 1991
This paper attempts to develop a theory of heuristics or preferences that can
be shared between understanding and generation systems. We first develop a
formal analysis of preferences and consider the relation between their uses in
generation and understanding. We then present a bi-directional algorithm for
applying them and examine typical heuristics for lexical choice, scope and
anaphora in more detail.
================================================================================
Title: Resolution of Collective-Distributive Ambiguity Using
Model-Based Reasoning
Author: Chinatsu Aone
Number: ACT-NL-117-91
Date: April, 1991
I present a semantic analysis of collective-distributive ambiguity, and
resolution of such ambiguity by model-based reasoning. This approach goes
beyond Scha and Stallard [12], whose reasoning capability was limited to
checking semantic types. My semantic analysis is based on Link [7, 8] and
Roberts [10], where distributivity comes uniformly from a quantificational
operator, either explicit (e.g., each) or implicit (e.g., D-operator). I
view the semantics module of the natural language system as a hypothesis
generator and the reasoner in the pragmatics module as a hypothesis filter
(c.f. Simmons and Davis [13]. The reasoner utilizes a model consisting of
domain-dependent constraints and domain-dependent axioms for disambiguation.
There are two kinds of constraints, type constraints and numerical
constraints, and they are associated with predicates in the knowledge base.
Whenever additional information is derived from the model, the Contradiction
Checker is invoked to detect any contradiction in a hypothesis using simple
mathematical knowledge. CDCL (Collective-Distributive Constraint Language) is
used to represent hypotheses, constraints, and axioms in the way isomorphic to
diagram representations of collective-distributive ambiguity.
================================================================================
Title: Knowledge-Based Natural Language Processing: the KBNL System
Author(s): J. Barnett, D. D'Souza, K. Knight, I. Mani, P. Martin,
E. Rich, C. Aone, J. Blevins, W. Bohrer, S. Luper-Foy, J.C.
Martinez, D. Newman
Number: ACT-NL-123-91
Date: April, 1991
In this paper we describe a knowledge-based natural language processing
system, KBNL. KBNL is novel in the way it is connected to backend knowledge
bases via a generic NL/KB interface, in the clean separation it enforces
between linguistic knowledge and world knowledge, in its use of knowledge to
aid in lexical acquisition, and in its exploitation of a single knowledge base
to support both language understanding and language generation. We describe
three applications of KBNL: text retrieval, machine translation, and the
combination of them--multilingual text retrieval.
================================================================================
For Digital Internal Use Only
April, 1991 MCC Technical Report Abstracts & Revisions
Title: Deductive Databases in Action
Author: Shalom Tsur
Number: STP-LD-002-91
Date: March, 1991
The theory of Deductive Databases has been a topic of extensive research over
the last years. A number of prototype systems that implement some of the
theoretical results are in various stages of development. Despite this
intense activity, not enough has been done so far to identify the areas of
application in which this emerging technology could be most fruitfully
deployed. This paper surveys some of these potential areas, comments on our
experience, and suggests some directions for further research that have
emerged in the course of our attempts to use this technology.
================================================================================
Title: Non-Determinism in Deductive Databases
Author(s): Fosca Giannotti, Dino Pedreschi, Domenico Sacca, Carlo Zaniolo
Number: STP-LD-003-91
Date: March, 1991
This paper examines the problem of adding non-deterministic constructs to a
declarative database language based on Horn Clause Logic. We revise a
previously proposed approach, the choice construct introduced by Krishnamurthy
and Naqvi, from the viewpoints of amenability to efficient implementation and
expressive power. Thus, we define a construct called dynamic choice, which is
consistent with the fixpoint-based semantics, fixes some deficiences of the
former approach, and leads to efficient implementations in the framework of
deductive databases. Also the new construct extends the expressive power of
Datalog programs considerably, as it allows to express negation under Closed
World Assumption, as well as a class of relevant deterministic problems.
================================================================================
Title: Minimum and Maximum Predicates in Deductive Databases
Author(s): Sumit Ganguly, Sergio Greco, Carlo Zaniolo
Number: STP-LD-124-91
Date: March, 1991
A novel approach is proposed for expressing and computing efficiently a large
class of problems, including finding the shortest path in a graph, that were
previously considered impervious to an efficient treatment in the declarative
framework of logic-based languages. Our approach is based on the use of min
and max predicates having a first-order semantics defined using rules with
negation in their bodies. We show that when certain monotonicity conditions
hold then (1) there exists a total well-founded model for these programs
containing negation, (2) this model can be computed efficiently using a
procedure called greedy fixpoint, and (3) the original program can be
rewritten into a more efficient one by pushing min and max predicates into
recursion. The greedy fixpoint evaluation of the program expressing the
shortest path problem coincides with Dijkstra's algorithm.
================================================================================
Title: The LDL Fact Manager
Author: Natraj Vidur Arni
Number: STP-LD-125-91(P)
Date: April 3, 1991
This is a description of the LDL fact manager. The fact manager provides the
run time database support for LDL and the LDL programming environment, SALAD.
This document describes the architecture of the fact manager and the data
model used. Each one of the functions used to implement the fact manager are
described in detail. A brief overview of the SALAD architecture is also
given.
================================================================================
Titles: LDL Tutorial, Rev. 1, and LDL Tutorial: Workbook
Date: March 20, 1991
================================================================================
Title: Overview and Status: Topics and Deliverables
Author(s): Susan Gerhart, Mark Bouler, Kevin Greene, Damir Jamsek,
Ted Ralston, David Russinoff
Number: STP-FT-001-91(P)
Date: March 1, 1991
The content of the FMTS (Formal Methods Transition Study) is described:
Methods, Tools, Process, Case Studies, Trends, Education, and Experiments.
Deliverables are defined with their current status. This document provides
the overall project view and should be read before the remaining specialized
reports. Each report contains both descriptions and assessments of the state
of theory and practice.
================================================================================
Title: State Transition Systems
Author(s): Mark Bouler, David Russinoff
Number: STP-FT-002-91(P)
Date: March 1, 1991
Specifying systems by their states and transitions is a widely-used approach
in various fields. This is formalized by the notion of state transition
systems. Many informal specifications of different kinds can be expressed by
such systems. Moreover, many specification and programming languages (and
their associated logics) can be seen as formalisms for specifying and
reasoning about such systems. Thus if a system can be specified as a state
transition system, one can use any of these formalisms for specification of or
reasoning about the system. The comparative advantages and disadvantages of
these formalisms for specifying different aspects of state transition systems
are discussed.
================================================================================
Title: Z: Notation & Use
Author(s): Kevin Greene, Mark Bouler, Damir Jamsek, Susan Gerhart,
Ted Ralston
Number: STP-FT-003-91(P)
Date: March 1, 1991
The Z notation is presented as one formal method for specifying software
systems. Included in this report is summary of the notation and its use
within an FMTS framework, a new graphical representation of Z specifications,
summaries of and pointers to some examples of Z specification that can be
found in the literature, and finally our assessment of Z.
================================================================================
Title: Object-Oriented Programming & Formal Methods
Author(s): Kevin Greene
Number: STP-FT-004-91(P)
Date: March 1, 1991
This preliminary version of the object-oriented programming and formal methods
(OOFM) report is essentially the result of transforming the OOFM presentation
made at the December 1990 FMTS workshop into a report. Some updates have been
made to existing projects as further information has been made available. In
addition to the projects reported on in December, several others have been
added to the survey: Eiffel, Concurrent Objects, and ENVISAGER. Brief
descriptions of these projects have been added to the outline.
================================================================================
Title: Tools Overview and Taxonomy
Author(s): Susan Gerhart, Kevin Greene
Number: STP-FT-005-91(P)
Date: March 1, 1991
A taxonomy of tools supporting Formal Methods is provided, including
descriptions and examples of each class of tools. The classes include:
editors, documentation aids, type and semantic analyzers, testing and proving
assistants, graphics, and object managers. A process description shows how
some of the tools may be used in context. Research issues are discussed.
================================================================================
Title: A Comparison of Theorem Proving Environments
Author(s): Damir A. Jamsek, David M. Russinoff
Number: STP-FT-006-91(P)
Date: March 1, 1991
This is a preliminary version of a report that will describe various theorem
proving environments. In this version, the HOL, NQTHM and Clio theorem
provers are presented, compared and contrasted from a user standpoint. A
brief assessment is made based on properties that seem to be important in
influencing the choice of theorem proving environment suitable for a
particular situation. In addition, a brief description of Larch, NUPRL and
Eves is given, as these three provers will complete the complement of provers
covered in the final report.
================================================================================
Title: Techniques for Assuring Correctness of Code
Author(s): Damir A. Jamsek, Kevin J. Greene
Number: STP-FT-007-91(P)
Date: March 1, 1991
In this preliminary version, various methods for assuring the correctness of
code are presented. Derivation of code from specification, transformational
techniques, mechanical reasoning, annotation methods, and practical
application of these methods is described. This preliminary report is meant
to provide an indication of the areas being investigated, brief assessment of
these techniques and a direction for the full report.
================================================================================
Title: The Role of Testing in Formal Methods
Author: Susan Gerhart
Number: STP-FT-008-91(P)
Date: March 1, 1991
Testing is the process of trying to refute the hypothesis that something is
correct. Various strategies for devising tests and assessing their ability to
refute correctness are described, most of which come from the research
literature. Specifications are viewed as both the object of test as well as
the basis for generating tests for implementations. One particular paper
offers a general model for testing during top-down design and a number of
interesting specific strategies worth following up. Suggestions for further
research are offered.
================================================================================
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.8 | FWD: MCC Systems Report | ULYSSE::ROACH | TANSTAAFL ! | Tue Jun 04 1991 12:21 | 142 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 29-May-1991 08:18pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD04@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC Systems Report
From: AUSTIN::KUHLMAN "Bill Kuhlman (512) 338-3243 29-May-1991 1244" 29-MAY-1991 12:52:44.66
To: @SYSNEWS
CC:
Subj: MCC Systems Report
MCC Systems Technology Report INTERNAL USE ONLY
*****************************
May 29, 1991 (113 lines)
Prepared by Bill Kuhlman and Ralph Cherubini, Austin, TX
This newsletter covers the work in the Object-Oriented and Distributed Systems
(OODS) lab within MCC's Advanced Computing Technology (ACT) Program and the
Deductive Computing Lab within the Software Technology Program (STP). Please
feel free to forward it within Digital. Address inquiries concerning Deductive
Computing projects to AUSTIN::CHERUBINI; OODS projects to AUSTIN::KUHLMAN.
Deductive Computing
===================
LDL
---
The MCC tutorial is available on request from RDVAX::MCC. Ask for "LDL Tutorial
Revision 1, March 20, 1991". Be sure to ask that you be sent a copy of the
accompanying workbook as well.
The objectives of the tutorial are to:
o Introduce deductive database technology
o Compare and contrast deductive database technology with current software
technologies
o Provide practice in writing and running LDL programs
Formal Methods Transition Study
===============================
There is no FM Today newsletter from MCC this month. Next month will include
a review of the FMTS workshop to be held mid-June.
OODS (Carnot)
=============
Project Review Meeting
----------------------
A review meeting of the Carnot project was held on May 7, 1991 at MCC.
Attending from Digital were Bob Taylor, TPWest, and Ken Beers, CRA. Funding
for Digital's participation in the project is being sought.
Technical Reports
-----------------
ACT-OODS-127-91 "Resource Integration Using an Existing Large Knowledge Base"
by Collet, et al. This report describes a method for integrating separately
developed information resources to enable them to be accessed and modified
coherently. It describes an evaluation of the method based on the
integration of three databases that have different data models (entity-
relationship, relational, and object-oriented) but similar semantics for
their data (i.e., the databases capture information about the same domain).
ACT-OODS-153-91(Q) "Specifying Interdatabase Dependencies in a Multidatabase
Environment" by Rusinkiewicz, et al. This paper describes a conceptual
framework for managing interdependent data. "Data Dependency Descriptors"
are used to specify data relationships between data objects. Together, the
Data Dependency Descriptors form an "Interdatabase Dependency Schema" that
can be used to control updates in a multidatabase environment.
Carnot Progress Update
----------------------
Distributed Transaction Environment via Rosette: Tested interoperability of
the SAG functionality with NCR and cleaned up errors in the ASN.1
implementation in Rosette.
SAG Interoperability via RDA: Completed and tested Rosette scripts for test D
of the SQL Access Group.
SAG OODB Server: Continued the implementation of an SQL query front end for
the Itasca-server. The Rosette RDA Service Agent for the Itasca-server is
complete enough to run the SAG interoperability Tests A, B, C, and D.
SAG Client Applications: Added to the Carnot Graphical Interaction Environment
the ability to use bitmaps for database objects and classes of objects.
Preparations for the July 1991 SAG Interoperability Demo: Continued to expand
the set of classes required by OOSI. Some effort was expended in implementing
abstractions for basic objects since GNU libg++ can't be used without falling
under the GNU Copyleft.
Graphical Interaction Environment (GIE) via HITS: Incorporating the motif
interface elements into Rosette.
Semantics-Based Graphical Languages: An implementation of a KD-Tree scheme
to store graphic data based on bounding boxes is complete; Beginning work on
the generation of entity-relationship diagrams from a description of a database
schema; Completed a rudimentary version of the line drawing (flowchart)
recognition/ parsing interface under X-windows (CLM).
Semantic Integration via CYC: Developed templates for constructing articulation
axioms, based on knowledge representation alternatives and common data model
structures; Established connections between schema integration module and
databases on different machines--Suns and Symbolics--through Lisp and Rosette;
Built a C interface for Ingres databases using ESQL; Constructed the schema
for an Orion database on automobiles for use in the CYC demonstration.
Graphical Articulation Axiom Generation Tool: Creating a Motif-based graphical
interface to the schema integration module of Carnot. The core of this is a
suite of tools for manipulating graphs.
===============================================================================
If you have suggestions for getting MCC technology into Digital or want to be
added to or removed from the distribution list, send mail to AUSTIN::KUHLMAN.
MCC technical reports are available through the Digital Library System or by
sending mail to RDVAX::MCC and including the TR number and name, your name,
badge number, cost center and mail-stop. Newsletters are also available on
three other MCC activities: Neural Networks, Artificial Intelligence and Optics.
Subscribe by sending mail to AUSTIN::KUHLMAN.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.9 | FWD: MAY AI 1991 MCC TECHNICAL REPORT ABSTRACTS & R | ULYSSE::ROACH | TANSTAAFL ! | Tue Jun 04 1991 14:46 | 70 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 03-Jun-1991 08:06pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD04@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MAY AI 1991 MCC TECHNICAL REPORT ABSTRACTS & REVISIONS
For Digital Internal Use Only
May, 1991 MCC Technical Report Abstracts & Revisions
Title: Treatment of Plurals and Collective-Distributive Ambiguity in
Natural Language Understanding
Author: Chinatsu Aone
Number: ACT-NL-155-91
Date: May, 1991
In this research, semantic theories of plural NPs and collective-distributive
ambiguity are developed, and a reasoning method for resolution of such
ambiguity in a natural language understanding system is studied. The
investigation of plural NPs begins with syntactic and semantic analyses of
measure constructions, which provide insight into the semantics of plural and
mass nouns. The syntax is formalized as a Unification-based Categorial
Grammar and the semantics is formalized using the Discourse Representation
Theory (DRT). Then, a semantic theory of collective-distributive ambiguity is
presented. It is based on the semantic theory that attributes distributivity
to quantificational operators. This theory is extended in the DRT framework
by introducing events into the representations. Finally, in order to resolve
collective-distributive ambiguity, a method using model-based reasoning is
proposed; it makes use of a model consisting of domain-dependent and
domain-independent knowledge to filter out inconsistent hypotheses. The
knowledge necessary for reasoning about collective-distributive ambiguity is
identified. Domain-dependent knowledge is represented as a set of constraints
and domain-independent knowledge is represented as a set of axioms in a typed
unification language called the "Collective-Distributive Constraint Language"
(CDCL).
================================================================================
Title: KBNL 1991 Spring Release: Installation Notes and User Guide
Author(s): J. Barnett, D. D'Souza, K. Knight, I. Mani, P. Martin
Number: ACT-NL-158-91-Q
Date: May, 1991
In this report we describe the latest release of the KBNL natural language
processing system. Included are instructions for installing the system and
for running both the text retrieval system Scan and the NL developer's
interface.
================================================================================
Title: KBNL 1991 Spring Release: Software Tape
Number: ACT-NL-161-91-Q
Date: May, 1991
================================================================================
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.10 | FWD: MCC AI Update | ULYSSE::ROACH | TANSTAAFL ! | Tue Jun 04 1991 14:47 | 126 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 28-May-1991 08:15pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD04@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC AI Update
MCC AI UPDATE No. 16 INTERNAL USE ONLY
********************
May 23, 1991 (102 lines)
Prepared by: Bill Kuhlman, MCC liaison, Austin, Texas
Introduction
============
AI Update attempts to inform the Digital community of activities and events
of interest in the MCC Artificial Intelligence Laboratory. Please feel free
to forward it within Digital. Additions and deletions to the distribution
may be sent to AUSTIN::KUHLMAN. To obtain copies of MCC technical reports
mentioned in the Update, send mail to RDVAX::MCC (include mail stop).
Upcoming Meetings
-----------------
June 27-28, 1991 First International Conference on Fuzzy Systems
Austin, TX
July 22-23, 1991 KBNL In-Depth, MCC, Austin, TX
KBNL (Knowledge-Based Natural Language) and Machine Translation
===============================================================
Spring software release
-----------------------
The new version of the KBNL system has been released. The system has been
tested on three platforms: Symbolics, Lucid Common Lisp (on Sun 3's and Sparc
Stations) and Allegro Common Lisp (on Sparc Stations). The port to the
DS3100 will proceed when more memory arrives.
ACT-NL-157-91-Q "KBNL 1991 Spring Release: System Documentation"
ACT-NL-158-91-Q "KBNL 1991 Spring Release: Installation Notes and User Guide"
ACT-NL-161-91-Q "KBNL 1991 Spring Release: Software Tape"
The development of the Garnet-based interface to the NL developer's toolkit has
continued with tests of the system on example texts. It is now very robust.
NL Tech Report
--------------
ACT-NL-155-91 "Treatment of Plurals and Collective-Distributive Ambiguity
in Natural Language Understanding" by C. Aone. This paper develops theories
of plural NPs and collective-distributive ambiguity. It studies a reasoning
method for resolution of such ambiguity in a NL understanding system.
The NL group finished four papers that will appear in this summer's round of
conferences: one at the Machine Translation Summit in Washington, one at the
Association for Computational Linguistics (ACL) annual meeting in Berkeley, and
two at an ACL workshop on reversible systems. These papers are part of the
strategy for positioning MCC for government funding of Machine Translation by
increasing visibility in the MT technical community.
CYC
===
The main efforts on the knowledge entering front this month were concerned with
representing human relations (such as a rule stating that family members are
all of the same species) and spousal and cohabitation relationships. They are
also working on assertions concerning desired features in a house, musical
instruments, hobbies, and professions.
The new portable (CLIM-based) interface platform (called "Kami"), designed to
support the MUE, UE, etc. on DS3100/5000s and Sun, was installed and work
progressed on getting all the new CLIM interface code running on both types
of DECstations
On the theoretical front, work was done on the semantics of microtheories and
its relationship to existing possible- world semantics. Other significant
theoretical work included working out more details concerning reasoning about
actions, using the 5-d space-time-intelligence ontology. Work began on new
explicit meta level inference guidance techniques. These may turn out to be
useful for guiding long inference chains, for selecting which microtheories to
import from in cases of conflicting choices, and for deciding among preference
criteria when choosing which arguments to prefer over which others.
Comings and Goings
-------------------
CYC held a project review meeting on May 7. It was attended by Phill Apley
from CRL and Ken Beers, CRA.
Chinatsu Aone, with the KBNL group, will visit DEC Japan RDC within the
next two weeks. Contact Yoshi Sato for more information.
================================================================================
Newsletters are also available on three other research areas at MCC:
Systems Technology, Neural Networks, and Optics. To subscribe, send mail to
AUSTIN::KUHLMAN. Reports from the MCC Experimental Systems project ES-Kit are
posted in the VAX Notes conference RDVAX::MCC-ES-KIT. For reports on MCC's
work on Packaging/Interconnect and High Temperature Superconductivity, contact
Chauvet Parker (AUSTIN::PARKER).
MCC is a cooperative enterprise whose mission is to strengthen and sustain U.S
and Canadian competitiveness in information technologies. The objective is
excellence in meeting broad industry needs through application-driven research,
development and timely deployment of innovative technology. The ACT (Advanced
Computing Technology) program mission is to anticipate and fulfill requirements
for computer systems technology that is five to ten years in advance of the
state-of-the-art, while maintaining a balanced R&D portfolio that provides
participants with continuing technology leadership. The theme of user-friendly,
distributed, heterogeneous systems is the basis for ACT's next five year vision.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.11 | FWD: KBNL Indepth, 7/22-23, MCC | ULYSSE::ROACH | TANSTAAFL ! | Tue Jun 11 1991 20:36 | 128 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 11-Jun-1991 07:16pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD03@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: KBNL Indepth, 7/22-23, MCC
From: AUSTIN::KUHLMAN "Bill Kuhlman (512) 338-3243 10-Jun-1991 1714" 10-JUN-1991 17:22:19.24
To: @MCC-AI
CC:
Subj: KBNL Indepth, 7/22-23, MCC
The Knowledge-Based Natural Language Project at MCC is planning an in-depth
technical review on July 22-23, 1991, in Austin. They would like to invite
those in Digital who might be interested to attend this review, which will
focus on this spring's release of a new KBNL system that is portable across
knowledge bases and that runs on standard platforms. The system now runs
on the DECstation. They will discuss both the core KBNL system (Lucy, Koko,
and Luke) as well as applications of the core system for text retrieval and
machine translation. They will be looking for feedback from participants
particularly on issues of how the system can be made more usable.
A preliminary agenda is attached. A more detailed agenda will be available
in a couple of weeks. If you plan to attend this meeting, PLEASE RETURN THE
ATTACHED TRAVEL REQUEST FORM TO AUSTIN::KUHLMAN.
Knowledge-Based Natural Language In-Depth Meeting
July 22-23, 1991
Microelectronics and Computer Technology Corporation
3500 West Balcones Center Drive
Austin, Texas 78759
Preliminary Agenda
Monday, July 22, 1991
8:30 Intro to KBNL for new people
9:30 Overview of current system
12:00 Lunch
1:00 Key issues: reversibility, KB independence, multilingual systems
3:30 Hands-on
6:30 Dinner
Tuesday, July 23, 1991
9:00 Applications
12:00 Lunch
1:00 Feedback discussion, future directions
3:00 Adjourn
T R A V E L R E Q U E S T F O R M
Name:
DTN:
Organization:
Cost Center:
CC Manager:
Travel itinerary:
Purpose of visit:
Who visiting:
=======================================================================
Only complete if requesting TTP funded travel.
Financial analyst:
DTN:
E-mail address:
Estimated Expense of trip:
=======================================================================
TECHNOLOGY TRANSFER PROGRAM OFFICE
This visit is approved/disapproved.
Trip report is due to RDVAX::MCC within 5 working days of travel.
__________________________ ____________
TTP Program Office Date
____________________________ ____________
Cost Center Manager Date
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.12 | FWD: Announcement, Alan Salisbury names COO of MCC | ULYSSE::ROACH | TANSTAAFL ! | Fri Jun 28 1991 11:04 | 85 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 27-Jun-1991 10:21pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD04@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: Announcement, Alan Salisbury names COO of MCC
From: CRAKJB::BEERS "Ken Beers DTN: 223-3694 27-Jun-1991 1447" 27-JUN-1991 14:14:22.22
To: @MCCDIS:MCC.DIS
CC:
Subj: Announcement, Alan Salisbury names COO of MCC
MCC NAMES ALAN SALISBURY EXECUTIVE VICE PRESIDENT
AND CHIEF OPERATING OFFICER
Austin, TX -- June 4, 1991 -- The Microelectronics and Computer
Technology Corporation (MCC) has announced that Dr. Alan B. Salisbury
has been named Executive Vice President and Chief Operating Officer
effective June 1. In that capacity, Salisbury will share responsibility
for the day-to-day operations of MCC with Dr. Craig Fields, President
and CEO, and Dr. Barry H. Whalen, Vice President, Marketing and Customer
Services. The three will work jointly to manage the implementation of
MCC's recently approved ten-year vision, as well as MCC's relations with
its supporting organizations.
"Alan has a demonstrated track record of success, and had particularly
excelled in the management and application of information technology and
research as corporate strategic weapons," according to Dr. Fields.
"Alan's proven ability to manage research and technology transfer, and
his extensive experience in information and communications management,
make him ideally suited to the task of operating MCC."
Prior to joining MCC, Salisbury was president of the Contel Technology
Center, the advanced technology research and development division
serving Contel Corporation. Prior to its acquisition earlier this year
by GTE, Contel was the nation's third largest independent telephone
company.
At the Contel Technology Center, Salisbury was responsible for
laboratories covering the full spectrum of telecommunications and
information systems technologies; software engineering, artificial
intelligence, networks and secure systems, and transmission and
switching systems.
Salisbury joined Contel in 1987, after retiring from 29 years in the
U.S. Army Signal Corps with the rank of Major General. In his final
military assignment, he served as Commander, U.S. Army Information
Systems Engineering Command, a 4500+ person, world-wide information
systems engineering and integration organization, responsible for design,
development, acquisition, and support of hundreds of management
information, telecommunications, and command and control systems.
Dr. Salisbury holds masters and PH.D. degrees in electrical engineering
and computer science from Stanford University, and is also a graduate of
the U.S. Military Academy. Dr. Salisbury has had extensive teaching
experience, and currently serves on the Boards of Visitors for the
Software Engineering Institute at Carnegie Mellon University and the
College of Engineering at the University of Maryland. He is a founding
editor of the "Journal of Systems and Software" and is the author of
numerous technical and management paper, as well as "Microprogrammable
Computer Architecture", a widely adopted computer architecture
reference.
MCC is a cooperative R&D enterprise whose mission is to strengthen and
sustain the competitiveness of member companies who share common
elements of a technical vision in information technology. MCC's
membership currently includes 22 shareholder and 36 associate members.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.13 | FWD: MCC Systems Report | ULYSSE::ROACH | TANSTAAFL ! | Wed Jul 17 1991 10:54 | 195 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 28-Jun-1991 06:09pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD05@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC Systems Report
MCC Systems Technology Report INTERNAL USE ONLY
*****************************
June 27, 1991 (171 lines)
Prepared by Bill Kuhlman, MCC liaison, Austin, TX
This newsletter covers the work in the Object-Oriented and Distributed Systems
(OODS) project within MCC's Advanced Computing Technology (ACT) Program and the
Formal Methods and Deductive Computing projects within the Software Technology
Program (STP). Please feel free to forward it within Digital.
Formal Methods Transition Study
===============================
Review Meeting Held
-------------------
The Formal Methods Transition Study held its second review meeting, June 18-19,
1991, at MCC. Attending from Digital was Gary Feldman, with the DECSpec
project in TLE. That is a joint project with MIT to develop a Larch
specification tool for the C language.
The meeting presented case studies, experiments, and assessments of various
methods: LOTOS, VDM, Larch, Gist, Z. Much of this information was released
to project participants in the following MCC technical reports:
STP-FT-194-91(P) "LOTOS" by K. Greene. This report describes LOTOS, one
of the four Formal Description Techniques (FDTs) standardized by the
telecommunications industry. LOTOS is a specification/design language
created for expressing the organization, communication, and behaviors
of a collection of concurrent processes comprising a system.
STP-FT-195-91(P) "VDM: Notation & Use" by M. Bouler. This paper presents
VDM as one formal method for specifying software systems. It includes a
summary of the notion and its use, summaries and pointers to the literature,
and an assessment. Comparisons are made to Z.
STP-FT-196-91(P) "Larch Method Assessment" by D. Jamsek. This report describes
the Larch Method of specification and verification. The specification has
two tiers: the Larch Shared Language and the Larch Interface Language.
The Larch Prover is the verification component of the system.
STP-FT-197-91(P) "Gist" by D. Russinoff. Gist is a specification language
designed for the development of composite systems consisting of several
interacting components. This report describes the general framework of
the language, illustrates the use of the associated tools, and offers as
assessment of both.
STP-FT-198-91(P) "Requirements Engineering with Formal Methods" by S.
Gerhart. This report describes how formal methods fit into the activity
of gathering information and objectives for a system to be developed.
It describes several approaches, conventional wisdom, and the current
state of knowledge.
STP-FT-199-91(P) "Case Studies: The Z Notation" by T. Ralston. This report
describes several cases in which the Z method has been used to specify
real industrial-grade applications.
STP-FT-200-91(P) "Trends in Formal Methods" by T. Ralston. This report
describes trends in standardization and industrialization, legal and
regulatory trends, and trends in formal methods education and training.
STP-FT-201-91(P) "Experiments: Preliminary Results" by Bouler et al.
This report presents preliminary results and plans for four FMTS
experiments.
STP-FT-148-91 "A Mechanically Verified Incremental Garbage Collector"
by D. Russinoff. As an application of a system designed for concurrent
program verification, this report describes a formalization and mechanical
proof of the correctness of Ben-Ari's incremental garbage collection
algorithm.
Plans for Follow-on
-------------------
A proposal for FMTS follow-on projects has been prepared. It is built in
three layers: Transition (FMTS continuation for new participants, including
Tools Tested and Education/Training), Technology Research (SpecTra integration
and experimentation), and Architecture (the collaboration proposal to DARPA).
OODS (Carnot)
=============
Carnot Technical Progress
-------------------------
Distributed Transaction Environment via Rosette: An exception processing
system for Rosette to support reporting of conditions over remote evaluation
and OSI connections was designed and implementation was begun. An improved
C-structure system for Rosette was designed. This facility is needed to support
the integration of complex subsystems such as Motif/X libraries and ISODE
libraries.
SAG/RDA-Client Extensible Services Switch (ESS): Prepared and executed a demo
of the SAG/RDA client interface. A general purpose interface the accepts
Rosette scripts for concurrent execution of SAG RDA SQL is ready for general
testing with SAG members.
SAG/RDA/OODB-Server Extensible Services Switch (ESS): Prepared and executed a
demo of the SAG/RDA Itasca server. The Itasca server is ready for general
testing by SAG members.
Preparations for the July 1991 SAG Interoperability Demo: requirements were met
for participation in the July '91 demonstration, so now the work really begins.
Graphical Interaction Environment (GIE) via HITS: GIE now handles cursor-based
drag and drop with a "widget under mouse" extension to CLM. The re-write of
GIE in Rosette is progressing. Work on full support for C-structs in Rosette
will begin soon.
Semantics-Based Graphical Languages: Initial designs have been worked out to
apply visual grammar representations and blackboard-based control technologies
to the problem of generating and updating graphical displays of database
schemas. Programmer's documentation was prepared for the KD-Tree graphical
object indexing system.
Semantic Integration via CYC: Prepared and executed a variation of the SAG
demo in which the distributed query processor, that uses Articulation Axiom
extracted from CYC, generated sub queries from a single initial SQL query and
retrieved data stored in several databases. Finished pre-integration tools
for both Relational (Ingres) Schema and Entity-Relational Schema.
Graphical Articulation Axiom Generation Tool: Building a Motif-based graphical
interface on top of the GIE common substrate to help users do schema
integration in CYC.
OODS Technical Reports
----------------------
ACT-OODS-208-91 "Unification-Based Grammars and Tabular Parsing for Graphical
Languages" by Wittenburg et al. This paper presents a unification-based
grammar formalism and parsing algorithm for the purposes of defining and
processing generalization of concatenative languages such as those found
in two-dimensional graphical domains.
ACT-OODS-214-91 "Resource Integration Without Application Modification" by
Shen et al. This paper describes a method for integrating separately
developed information resources that ensures that: 1) existing applications
remain intact, 2) the applications are usable as before, but simply have
access to the new information, and 3) they need only to be extended, not
altered, in order to make use of the new information.
Deductive Computing
===================
LDL++
-----
The design for the LDL++ abstract machine has been revised and documented.
Improved compilation techniques were devised for conditionals, recursion and
`forever' constructs. A novel architecture for supporting Abstract Data Types
(ADTs) has been designed: this combines ADT concepts from both the programming
languages and the database world.
Implementation of the new LDL++ system is progressing. A new LDL++ fact
manager is nearly completed and a new parser has been implemented.
Comings and Goings
------------------
Carlo Zaniolo, Natraj Arni and S. Tsur spent May 28 with Digital's Database
Systems Research group in Colorado Springs, reviewing the design and
implemenation plans for LDL++.
Roger Nasr, Digital assignee to the DC project, has completed his assignment at
MCC. He and Maria Nasr, who has been working in the Carnot project, are
scheduled to rejoin DEC, in New England, in July.
===============================================================================
Additions or deletions to the distribution should be sent to AUSTIN::KUHLMAN.
MCC technical reports are available through the Digital Library System or by
sending mail to RDVAX::MCC and including the TR number and name, your name,
badge number, cost center and mail-stop. Newsletters are also available on
three other MCC activities: Neural Networks, Artificial Intelligence and Optics.
Subscribe by sending mail to AUSTIN::KUHLMAN.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.14 | FWD: June AI 1991 MCC Technical Report Abstracts & | ULYSSE::ROACH | TANSTAAFL ! | Wed Jul 17 1991 10:55 | 80 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 28-Jun-1991 06:09pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD03@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: June AI 1991 MCC Technical Report Abstracts & Revisions
For Digital Internal Use Only
June, 1991 Technical Report Abstracts & Revisions
Title: Capturing Language-Specific Semantic Distinctions in
Interlingua-Based MT
Author(s): J. Barnett, I. Mani, E. Rich, C. Aone, K. Knight, J.C.
Martinez
Number: ACT-NL-003-91 (Rev 1)
Date: June, 1991
We describe an interlingua-based approach to machine translation, in which a
DRS representation of the source text is used as the interlingua
representation. A target DRS is then created and used to construct the target
text. We describe several advantages of this level of representation. We
also argue that problems of translation mismatch and divergence should
properly be viewed not as translation problems per se but rather as generation
problems, although the source text can be used to guide the target generator.
The system we have built relies exclusively on monolingual linguistic
descriptions that are also, for the most part, bi-directional.
==============================================================================
Title: Reversible Machine Translation: What to Do When the Languages
Don't Line Up
Author(s): Jim Barnett, Inderject Mani, Paul Martin, Eliane Rich
Number: ACT-NL-068-91 (Rev 1)
Date: June, 1991
In this paper we deal with issues that face an interlingua-based, reversible
machine translation system when the literal meaning of the source text is not
identical to the literal meaning of the natural target translation. We
present an algorithm for lexical choice that handles such cases and that
relies exclusively on reversible, monolingual linguistic descriptions and a
language-independent domain knowledge base.
===============================================================================
Title: A Minimal Encoding Approach to Feature Discovery
Author(s): Mark A. Derthick
Number: ACT-CYC-212-91
Date: June, 1991
This paper discusses unsupervised learning of orthogonal concepts on
relational data. Relational predicates, while formally equivalent to the
features of the concept-learning literature, are not a good basis for defining
concepts. Hence the current task demands a much larger search space than
traditional concept learning algorithms, the sort of space explored by
connectionist algorithms. However the intended application, using the
discovered concepts in the Cyc knowledge base, requires that the concepts be
interpretable by a human, an ability not yet realized with connectionist
algorithms. Interpretability is aided by including a characterization of
simplicity in the evaluation function. For Hinton's Family Relations data,
we do find cleaner, more intuitive features. Yet when the solutions are not
known in advance, the difficulty of interpreting even features meeting the
simplicity criteria calls into question the usefulness of any reformulation
algorithm that creates radically new primitives in a knowlege-based setting.
At the least, much more sophisticated explanation tools are needed.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.15 | FWD: MCC AI Update | ULYSSE::ROACH | TANSTAAFL ! | Wed Jul 17 1991 12:21 | 225 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 15-Jul-1991 03:52pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD03@DPD
Dept:
Tel No:
TO: See Below
Subject: FWD: MCC AI Update
MCC AI UPDATE No. 17 INTERNAL USE ONLY
********************
June 28, 1991 (202 lines)
Prepared by: Bill Kuhlman, MCC liaison, Austin, Texas
AI Update attempts to inform the Digital community of activities and events
of interest in the MCC Artificial Intelligence projects. Please feel free
to forward it within Digital. Additions and deletions to the distribution
may be sent to AUSTIN::KUHLMAN. To obtain copies of MCC technical reports
mentioned in the Update, send mail to RDVAX::MCC (include mail stop).
KBNL (Knowledge-Based Natural Language) and Machine Translation
===============================================================
KBNL Ported to DECStation
-------------------------
The new KBNL system is now running on the DECStation 3100 under Ultrix at MCC.
It has also been successfully installed at the JRDC in Japan where they plan to
start an evaluation of the LUCY English front-end system as a general tool.
We plan to install the sources and binaries on the net so that those interested
in it may obtain a copy. Pointers to the software will be included in the
next issue of this newsletter. This version does not now include the Luke
lexical acquisition tool, but they hope to have it by the time of the
indepth meeting.
KBNL Indepth Meeting - July 22 and 23, 1991
-------------------------------------------
The KBNL project will host an indepth technical review on July 22 and 23. If
you have in mind ways that natural language technology could be used in your
work, you are encouraged you to attend this meeting, both to find out more
about KBNL and to provide input to MCC. The goals of the meeting are to:
o Present enough detail on the current KBNL system (both through technical
presentations and via guided hands-on sessions) to enable participants to
use the system in their own work.
o Find specific applications that participants are working on and articulate
the best ways for MCC to help in getting those applications going.
o Provide guidance to the KBNL group on what your most important issues are.
This will help them to focus efforts over the next year in those areas that
will have the greatest impact on your work.
If you plan to attend, contact AUSTIN::KUHLMAN by July 8, or Linda Mitchell at
(512) 338-3705 (CRL::"[email protected]") thereafter, but in any case not later
than July 17. If you have specific issues in mind that you would like to make
sure get covered during the meeting, contact Elaine Rich at MCC.
Natural Language Strategy
-------------------------
As mentioned last month, a natural language strategy meeting was held in May.
The goal of the meeting was to bring together the companies that are currently
participating in MCC's natural language work, other MCC shareholders who have
expressed an interest in this area, and representatives from government
organizations that have needs for natural language technology. People from
Digital, Apple, Boeing, E Systems, Eastman Kodak, Motorola, NCR, Southwestern
Bell, U.S. West, DARPA, CIA, and DOD attended the meeting. The question people
were asked was, "What is the strategic role of natural language technology in
your organization and what are most important things for MCC to do in this
area?"
The responses to this question all centered around the role of natural language
(NL) as part of an effective interface, particularly to complex information
systems. In particular, the focus was on possible applications of knowledge-
based NL systems in which a knowledge base is used as a basis for representing
the meaning of linguistic expressions and for reasoning about those meanings.
Specific application areas that are of concern to many of the people at the
meeting were: interfaces to distributed databases (where the size and the
complexity of the set of possible queries make other kinds of query languages
inadequate), text retrieval (where the goal is to improve both accuracy and
recall over the rates that we see today in systems that ignore meaning), speech
and character recognition (where natural language can be used both to make
predictions that drive the front-end recognition process and to interpret the
output of that process), and interfaces to complex knowledge bases. There is
a consensus that the short term payoff for these applications will be systems
that work in English, but that over the mid to long term, a multilingual
capability is important. So machine translation is another important
application but not just as a stand alone process that translates volumes of
material. It is viewed primarily as a way to make other information systems
more effective. Spanish and Japanese are the highest priority, with a
significant gap between them and other languages, such as the other European
languages or Arabic, that are also of interest to some organizations.
Despite this emphasis on the use of natural language as part of larger
application systems, there was agreement that the major role that MCC should
play is not in the development of those systems. Instead, MCC needs to focus
its effort in two areas. The main one is the development of the core NL
technology. And the second is in working with the Members in helping them to
build applications that exploit that technology. There was a strong sentiment
expressed by the current Members that collaborative application development,
driven by the Members to meet their specific needs, is a key factor both in
directing the course of research on the core system and in making technology
transfer actually happen.
Technical Reports
-----------------
ACT-NL-003-91 (Rev.1) "Capturing Language-Specific Semantic Distinctions in
Interlingua-Based MT" by Barnett et al. This paper describes an interlingua
based approach to machine translation, in which a DRS representation of the
source text is used as the interlingua representation.
ACT-NL-068-91 (Rev.1) "Reversible Machine Translation: What to Do When the
Languages Don't Line Up" by Barnett et al. This paper deals with issues
that face an interlingua-based, reversible machine translation system when
the literal meaning of the source text is not identical to the literal
meaning of the natural target translation.
ACT-OODS-208-91 "Unification-Based Grammars and Tabular Parsing for Graphical
Languages" by Wittenburg et al. This paper presents a unification-based
grammar formalism and parsing algorithm for the purposes of defining and
processing generalization of concatenative languages such as those found
in two-dimensional graphical domains.
Coming & Goings
---------------
Chinatsu Aone, with the MCC KBNL group visited Digital's Japan R&D Center on
May 31. She gave an overview talk, a "tutorial" talk, and a demonstration of
the KBNL system.
Tom Wallis, manager of the newly formed Natural Language Program in Digital's
New Ventures, visited the KBNL project in Austin on June 26.
CYC (Large Scale Common Sense Knowledge Base
============================================
Microsoft joins CYC
-------------------
A contract from Microsoft was received on June 28, making them a new Level 1
member of the CYC project. A press release is expected to be issued shortly.
CYC Application Demo
--------------------
A sample application of CYC has been developed and demonstrated. It involves
helping a user select which type of automobile to purchase. The differences
between this and a standard expert system involve the catching of errors, based
on general knowledge, in case the user provides nonsensical or contradictory
information to the system (e.g., due to mistyping some information,
misunderstanding a question, etc.). It also exhibits successful integration of
several data bases, including commercial ones such as the Kelly Blue Book DB,
having varied schemas and even types (relational, object-oriented, etc.).
Rules have also been written for a house-purchase-selection demo. Unlike
the automobile demo, the house one involves selecting from a set of individual
homes on the market, rather than from a set of abstract `types.' The interface
will also be quite different, and the knowledge is completely contained within
the CYC KB, not in external DBs.
Coding
------
Time has been spent in deleting CYC units which are so incomplete and/or
contradictory that the system is better off without them. Most of these were
entered years ago, in some cases as much as five years ago, before the ontology
or representation language stabilized. Some new consistency-checking and
clean-up code was written, to help police the knowledge base. To do that in a
more principled and general fashion, a new automatic `self-test' subsystem is
being designed and is expected to be finished in time for the upcoming CYC
software release scheduled for September 15th.
Port to DEC Platform
--------------------
Work continued on the CLIM-based CYC editor (KAMI) and other interface tools,
both under Genera and Lucid CommonLisp, on the DECStation 5000.
CYC Technical Report
--------------------
ACT-CYC-212-91 "A Minimal Encoding Approach to Feature Discovery" by M.
Derthick. This paper is an unabridged version of ACT-CYC-234-90, "The
Minimum Description Length Principle Applied to Feature Learning and
Analogical Mapping". It discusses unsupervised learning of orthogonal
concepts on relational data.
MCC Executive Management Changes
================================
Les Belady, Vice President of MCC's Advanced Computing Technology (ACT)
Program and the Software Technology Program (STP) retired from MCC at the end
of May. Alan Salisbury was named Executive Vice President and Chief Operating
Officer effective June 1. Alan holds a doctorate from Stanford in EE and
Computer Science and he was most recently President of the Contel Technology
Center. Alan will serve as Acting Director of STP. Ron Riedesel, ACT
marketing manager, will serve as acting director of ACT.
================================================================================
Newsletters are also available on three other research areas at MCC:
Systems Technology, Neural Networks, and Optics. To subscribe, send mail to
AUSTIN::KUHLMAN. For reports on MCC's work on Packaging/Interconnect and High
Temperature Superconductivity, contact Chauvet Parker (AUSTIN::PARKER).
MCC is a cooperative enterprise whose mission is to strengthen and sustain
Noth American competitiveness in information technologies. The objective is
excellence in meeting broad industry needs through application-driven research,
development and timely deployment of innovative technology. The ACT (Advanced
Computing Technology) program mission is to anticipate and fulfill requirements
for computer systems technology that is five to ten years in advance of the
state-of-the-art, while maintaining a balanced R&D portfolio that provides
participants with continuing technology leadership. The theme of user-friendly,
distributed, heterogeneous systems is the basis for ACT's next five year vision.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO Ed Hurry@DVO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.16 | Fwd: MCC Systems Technology Report | ULYSSE::ROACH | TANSTAAFL ! | Mon Aug 12 1991 19:03 | 129 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 02-Aug-1991 10:16pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD03@DPD
Dept:
Tel No:
TO: See Below
Subject: Fwd: MCC Systems Technology Report
MCC Systems Technology Report INTERNAL USE ONLY
*****************************
July 30, 1991 (104 lines)
This newsletter covers the work in the Object-Oriented and Distributed Systems
(OODS) project within MCC's Advanced Computing Technology (ACT) Program and the
Formal Methods and Deductive Computing projects within the Software Technology
Program (STP). Please feel free to forward it within Digital.
OODS (Carnot)
=============
Carnot/SQL Interoperability Demo - A Success
--------------------------------------------
MCC's Carnot technology was part of the successful SQL Access Group (SAG)
demo in New York on July 16. For the first time, SAG - a consortium of 42
leading database software and hardware companies - demonstrated multi-vendor
interoperability using SQL as a standard query language.
The demo showed tools from different vendors talking to a number of other
vendors' databases in a simulated real-world environment where multiple
clients and servers exchanged information. Participating as clients in
the demo were ASK/Ingres, Cincom Systems, Digital, HP, Informix, Microsoft,
Oracle, Sun and Unify. On the server side were Digital, Fujitsu, HP, Oracle,
NCR, Tandem, Teradata and Unisys. Digital's DECquery and Rdb participated
as a client and server, respectively. Carnot also participated as both a
client and server, with the Itasca object-oriented database as a server
and the Graphical Interaction Environment (GIE) as client. GIE is based
on MCC's HITS Human Interface technology which has been taken off of the
Symbolics environment and now runs on a DECstation 5000 under Unix.
Patent Filed
-------------
A patent covering the ESS technology was filed on June 21, 1991. The patent
was filed before the code was publicly displayed so that patent applications
can now be filed anywhere in the world (especially Japan).
The Carnot Membership/Funding Situation
---------------------------------------
Digital is not a member of the OODS/Carnot project. For the past several months
we have been evaluating the technology. Given the current climate, membership
is not expected. However, MCC has received adequate funding to proceed with the
project (current funders are Bellcore, Itasca, NCR, and Eastman Kodak, with
Andersen Consulting and NSA scheduled to join later this year). This
newsletter will strive to report any major technical developments.
OODS Non-confidential Technical Reports
---------------------------------------
ACT-OODS-176-91 "Towards a Formal Theory of Communication for Multiagent
Systems" by M. Singh. This paper provides a formal semantics for the major
kinds of speech acts. The resulting theory applies uniformly to a wide range
of multiagent systems. Some applications of this theory are outlined, and
some of its theorems listed.
ACT-OODS-177-91 "A Logic of Situated Know-how" by M. Singh. This paper
presents sound and complete axiomatizations for two non-reductive and
intuitively natural formal definitions of the know-how of an agent situated
in a complex environment.
Formal Methods
==============
The one-year Formal Methods Transition Study is scheduled to be completed
in September. At that time, they will issue a new edition of the "FM Today"
newsletter.
On July 23, a project definition meeting was held at MCC to unveil initial
plans for a 3-year formal methods technology project, starting Oct 1, 1991.
The purpose of the meeting was to define the project objectives, from which
will follow a specific research program and deliverables plan for approval
in August.
Deductive Computing
===================
LDL
---
The implementation of a new user interface, LDL++ parser, predicate connection
graph extractor, and recursion rewriter is completed. Implementation of the
fact manager and a enhanced PCG interpreter is approaching completion.
Comings and Goings
==================
Carlo Zaniolo, and S. Tsur, with the Deductive Computing project, visited
Digital's Database Systems Research group in Munch, Germany in July. They
reviewed the design and implementation plans for LDL++ with C. Freytag, and
discussed possible deployment of the system in support for manufacturing
applications (Express), with Reiner Reschke.
Phil Cannata, Director of the Carnot project, visited Dieter Gawlick, with
the Activity Management Group in Mountain View, in June.
Steve Forgey, with the office group in TNSG, attended the Extensible Software
Platform workshop at MCC, July 15-17.
===============================================================================
Additions or deletions to the distribution should be sent to AUSTIN::KUHLMAN.
MCC technical reports are available through the Digital Library System or by
sending mail to RDVAX::MCC and including the TR number and name, your name,
badge number, cost center and mail-stop. Newsletters are also available on
three other MCC activities: Neural Networks, Artificial Intelligence and Optics.
Subscribe by sending mail to AUSTIN::KUHLMAN.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO DONALD HEIDORN@DYO
Ed Hurry@DVO yoshinori ishii@TKO phil james@DLB norio2 murakami@TKO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.17 | Fwd: July 1991 MCC STP Technical Report Abstracts | ULYSSE::ROACH | TANSTAAFL ! | Mon Aug 12 1991 19:03 | 170 |
|
I N T E R O F F I C E M E M O R A N D U M
Date: 02-Aug-1991 10:30pm CET
From: BEANE
BEANE@BIGRED@MRGATE@DPD03@DPD
Dept:
Tel No:
TO: See Below
Subject: Fwd: July 1991 MCC STP Technical Report Abstracts
From: HSOMAI::RDVAX::MCC "01-Aug-1991 1423" 1-AUG-1991 13:38:41.72
To: @DIST:STP
CC: MCC
Subj: JULY 1991 STP MCC Technical Report Abstracts & Revisions
For Digital Internal Use Only
July, 1991 MCC Technical Report Abstracts and Revisions
Title: A Mechanically Verified Incremental Garbage Collector
Author(s): David M. Russinoff
Number: STP-FT-148-91
Date: April, 1991
As an application of a system designed for concurrent program verification, we
describe a formalization and mechanical proof of the correctness of Ben-Ari's
incremental garbage collection algorithm. This system is based on the
Manna-Pnueli model of concurrency and implemented as an extension of the
Boyer-Moore prover. The correctness of the garbage collector is represented
by two theorems, stating (a) that nothing except garbage is ever collected
(safety), and (b) that all garbage is eventually collected. We compare our
formal treatment with several published informal proofs of the same results.
===============================================================================
Title: LOTOS
Author(s): Kevin J. Greene
Number : STP-FT-194-91 (P)
Date: May 29, 1991
LOTOS, one of the four Formal Description Techniques (FTDs) standardized by
the telecommunications industry, is described herein. LOTOS is a
specification/design language created for expressing the organization of,
communication between, and individual behaviors of a collection of concurrent
processes comprising a system. LOTOS may be used as a "pure" process algebra
or it may be used as a "pure" abstract data type (ADT) defining language.
What make LOTOS unique (and specifications written in it difficult to reason
about) is that these relatively well-understood paradigms may be mixed in a
single LOTOS specification. The design of LOTOS is based on many of the ideas
present in the process algebras CCS and CSP and on the ADT concepts found in
the specification language ACT ONE. LOTOS was given a formal semantics based
on structured labelled transition systems in the 1988 ISO document ISO
8807[16].
===============================================================================
Title: VDM: Notation & Use
Author(s): Mark Bouler
Number: STP-FT-195-91 (P)
Date: May 29, 1991
The VDM notation is presented as one formal method for specifying software
systems. Included in this report is a summary of the notation and its use
within an FMTS framework, summaries of and pointers to some examples of VDM
specification that can be found in the literature, and finally our assessment
of VDM. Comparisons are made throughout to Z, the formalism we feel most
clearly invites camparison with VDM.
===============================================================================
Title: Larch Method Assessment
Author(s): Damir A. Jamsek
Number: STP-FT-196-91 (P)
Date: May 29, 1991
This report describes the Larch Method of specification and verification. The
specification component of the method has two tiers. The first tier, an
algebraic approach to specification called the Larch Shared Language, forms
the basis for programming language specific components expressed in Larch
Interface Languages, the second tier. The semantics for both these components
is based on the logic of the Larch Prover which is the verification component
of the overall system. The three components are presented via examples and
an assessment of the method is given.
==============================================================================
Title: Gist
Author(s): David Russinoff
Number: STP-FT-197-91 (P)
Date: May 29, 1991
Gist is a specification language designed for the development of composite
systems consisting of several interacting components. It is intended to
support a closed-system specification style, in which the development of
components proceeds from a description of the behavior of a complete system,
without reference to any interaction with an unspecified external environment.
The language combines several computational paradigms, including frame-based
reasoning, a state-transition model, and constraint reasoning. It is
supported by a set of tools intending to help a user to create, debug,
validate, and explain a specification. This report describes the general
framework of the language, illustrates the use of the associated tools, and
offers an assessment of both.
================================================================================
Title: Requirements Engineering with Formal Methods
Author(s): Susan Gerhart
Number: STP-FT-198-91 (P)
Date: May 29, 1991
Requirements Engineering is the activity of gathering information and
objectives for a system to be developed. This report describes how formal
methods fit into this activity by surveying several approaches, compiling some
conventional wisdom about the nature of the activity, and assessing the
current state of knowledge and opportunities for greater use of formal
methods.
================================================================================
Title: Case Studies: The Z Notation
Author(s): Ted Ralston
Number: STP-FT-199-91 (P)
Date: May 29, 1991
This section of the FMTS contains descriptions of several cases in which the Z
method has been used to specify real industrial -grade applications. The
cases presented are: The IBM Customer Information Control System (CICS);
development on an SSADM toolset by Praxis Systems Ltd.; the use of Z by
Tektronix to develop a series of oscilloscopes; and the specification of the
control system of a clinical cyclotron system in the Neutron Radiation Therapy
Center at the University of Washington.
===============================================================================
Title: Trends in Formal Methods
Author(s): Ted Ralston
Number: STP-FT-200-91 (P)
Date: May 29, 1991
This section of the FMTS describes various trends in Formal Methods, namely,
trends in standardization and industrialization, legal, and regulatory trends,
and trends in formal methods education and training. General information
about the pace and uptake of Formal Methods by industry is presented,
referencing the section on Case Studies for more detailed descriptions of
actualc ases. Specifics on both the development of standards requiring Formal
Methods, as well as the use of Formal Methods in standard is provided. Several
legal cases are discussed, together with analysis of significant law review
articles on liability in computer law, and efforts by various regulatory
bodies worldwide are described. Finally, trends in education and training
courses in Formal Methods are presented along with an inventory of courses in
the various methods.
===============================================================================
Title: Experiments: Preliminary Results
Author(s): M. Bouler, S.L. Gerhart, K.J. Greene, D.A. Jamsek, T.J.
Ralston, D. Russinoff, J. Srinavasan
Number: STP-FT-201-91 (P)
Date: May 29, 1991
Preliminary results obtained and plans for each of the four FMTS Experiments
are presented.
Distribution:
TO:
DENNIS DICKERSON@DL STEVE DONOVAN@DLO Tommy Gaut@HSO DONALD HEIDORN@DYO
Ed Hurry@DVO yoshinori ishii@TKO phil james@DLB norio2 murakami@TKO
Louis Pau@VBE jim rather@HSO Pat Roach@VBE Czarena Siebert@HSO
Mike Sievers@HSO Dale Stout@HSO Susan Sugar@MWO Sherry Williams@HSO
Mike Willis@HSO Tom Wilson@HST
|
| 310.18 | Fwd: MCC day at the Mill (21 Nov 1991) | MR4DEC::ROACH | TANSTAAFL ! | Tue Nov 12 1991 20:52 | 112 |
|
I N T E R O F F I C E M E M O R A N D U M
Doc. No: 006396
Date: 12-Nov-1991 12:44pm EST
From: BEANE
BEANE@SHLACT@SELECT@MRGATE@NRGATE@NRO
Dept:
Tel No:
TO: PATRICK ROACH@LMO
Subject: Fwd: MCC day at the Mill (21 Nov 1991)
From: HSOMAI::RDVAX::MCC "MCC Program Office - DTN: 223-2069 05-Nov-1991 1317" 5-NOV-1991 14:00:22.27
To: @DIST:MCC.DIS
CC: MCC
Subj: Announcement of MCC Day - November 21, 1991
On November 21, 1991 from 8:30-5:00 in the General Doriot Auditorium in
the Mill, there will be a series of presentations dealing with new
technologies originating from MCC, the research consortia located in
Austin, Texas. Key researchers will be presenting overviews of their
projects that are of potential interest to Digital. An agenda will be
available shortly, please forward to any parties that might be interested in
these presentations. If you have any questions, please contact Ken Beers in
the Technology Transfer Program Office (DTN: 223-3694)
The subjects that will be covered will be:
International Liaison Office and Membership Services - An
overview of the services that are available to all shareholders,
including the tracking of world-wide sources of technologies (with a
focus on Japanese and European research)
Enterprise Integration - MCC has developed a program that will enable
companies to more effectively address these critical competitive requirements.
Focused on multi-organization and multi-company business relationships, the EI
Program is designed to perform work in real-world settings, emphasize near
term benefits to participants, and leverage the work of other consortia,
universities, member companies, and technology start-ups. With MCC carefully
cross fertilizing project efforts, technology will be re-used, and the
resulting solutions will be inter-operable. Participating companies will
benefit by having access to both the technologies and the multi-company
partnerships that help to form the basis for a national Enterprise Integration
network.
Power Sources (Batteries) - The industry trend toward more powerful
computers, higher resolution displays, and greater portability creates the
need for higher energy density and lighter weight power systems. To identify
candidate systems, MCC has initiated a review of electrochemical power sources
for portable electronics (such as laptop and notebook computers) and is
forming a research program to develop a series of very lightweight, reliable,
high power/density, solid state battery technologies. MCC is focusing on
lithium systems, and the most promising of these appears to be the lithium
cell (lithium doped polymer -- V60l3 or TiS2), using thin film technologies.
Packaging for Portable Electronics - Recent press articles speculate that the
major economic battle in the electronics industry will be decided by success
in manufacturing palmtop devices. In this project, methods of system packaging
used in cameras and hand held devices will be analyzed and then combined with
new technologies to reduce the size, weight, and cost of other portable
electronic devices such as notebook/palmtop/laptop computers and telephones.
Close interaction with vendors will ensure sources of supply for the developed
technologies.
Mirage - Advances in hardware are now making highly dynamic 2D and 3D
visual interfaces economically feasible. The Mirage Project is an effort
to substantially reduce the programming time and expertise required to
produce dynamic visual applications. Mirage aims to provide a framework
for developing and delivering interactive visual applications that are cost
effective through the unification of windowing, graphics, animation,
interaction, and simulation technologies within a heterogeneous, distributed
environment. By dramatically reducing the complexity and the amount of
expertise required to produce animated visualizations
Fuzzy Systems - MCC is launching a new project in the area of Fuzzy Systems.
While Japan has fielded thousands of Fuzzy applications, there has been
limited activity in the U.S. It has become clear that the two primary
roadblocks to the utilization of the technology by the broad market are the
isolation of existing Fuzzy tools from mainstream computer systems and the
lack of industry (or de facto) standards. The focus of the CRAFT project is
on bringing Fuzzy technology into main- stream computing. In addition, the
CRAFT project will assist in the development of international standards for
Fuzzy Systems and in the imple-mentation of prototype software that conforms
to those standards. This work will be based on and reinforced by
customer-supplied applications.
Optics (Bobcat) - This project is developing technologies for using laser
optics to store digital information as holograms in photorefractive crystals.
This type of storage would allow access and transfer speeds that are 100 to
1,000 times faster than magnetic disk. It promises to be much less expensive
than DRAM and is non-volatile (no power required to refresh). Since it has no
moving parts, holographic storage is potentially more reliable and
maintainable than traditional magnetic media. The project has collaboration
agreements with Stanford University, the University of Colorado, and the
University of Rochester.
CAD Framework - The CFL exists to support the requirements of its
participants with regard to CAD framework standardization. The primary
objective is to support the technical activities of the CAD Framework
Initiative (CFI), an industry-supported group seeking to define interface
guidelines between CAD tools and CAD frameworks in order to reduce or to
remove barriers to tool/framework integration. The CFL also provides
proprietary services to its members, including review of draft CFI interface
guideline proposals, technical support for becoming compliant with CFI
guidelines, and training/education in the use of CFI guidelines. In addition
to commercial participants, the CFL receives funding from the Defense Advanced
Research Projects Agency (DARPA)
|
| 310.19 | YES | MR4DEC::ROACH | TANSTAAFL ! | Thu Nov 14 1991 17:21 | 49 |
|
I N T E R O F F I C E M E M O R A N D U M
Doc. No: 006451
Date: 14-Nov-1991 10:59am EST
From: BEANE
BEANE@SHLACT@SELECT@MRGATE@NRGATE@NRO
Dept:
Tel No:
TO: PATRICK ROACH@LMO
Subject: FWD: (Fwd: MCC) October 1991 ACT MCC Technical Report Abstra
From: RDVAX::MCC "MCC Program Office - DTN: 223-2069 04-Nov-1991 1517"
4-NOV-1991 14:25:02.04
To: @DIST:ACT
CC: MCC
Subj: October 1991 ACT MCC Technical Report Abstracts & Revisions
For Digital Internal Use Only
October 1991 Technical Report Abstracts & Revisions
Title: Visual Metaphors: 3D Interfaces for Knowledge Workers
Author(s): Steven E. Poltrock, John L. Lovgren, Mark A. Tarlton, P. Nong
Tarlton
Number: ACT-348-91
Date: September 1991
Between 1984 and 1987 a team of researchers in the Human Interface Program at
the Microelectronics and Computer Technology Corporation (MCC) investigated
innovative ways of building and using 3D user interfaces and graphic realism
to support knowledge workers. Our research included knowledge-based
applications of 3D user interfaces such as SemNet and was principally focused
on a user interface development environment we called the Metaphor Builder.
This suite of tools included a representation scheme called GrafBag, a 3D
editor called Visage, and a model editor called GrabGraf. We validated these
tools by building the Visual Office, a 3D environment for managing work
contexts.
This research was accomplished in an advance development created by
integrating a Lisp machine with a 3D display engine. We describe how the
Visual Office was designed and implemented using components of the Metaphor
Builder. The current Mirage project at MC traces its roots to this research.
|
| 310.20 | Fwd: MCC AI Update #20 | MR4DEC::ROACH | TANSTAAFL ! | Thu Dec 05 1991 16:02 | 198 |
|
I N T E R O F F I C E M E M O R A N D U M
Doc. No: 006813
Date: 04-Dec-1991 10:17pm EST
From: BEANE
BEANE@SHLACT@SELECT@MRGATE@NRGATE@NRO
Dept:
Tel No:
TO: PATRICK ROACH@LMO
Subject: Fwd: MCC AI Update #20
From: BIGRED::AUSTIN::KUHLMAN "Bill Kuhlman (512) 338-3243 04-Dec-1991 1655" 4-DEC-1991 17:07:04.63
To: @MCC-AI
CC:
Subj: MCC AI Update
MCC AI UPDATE No. 20 INTERNAL USE ONLY
********************
December 4, 1991 (176 lines)
AI Update covers MCC's CYC project (Large Scale Common-Sense Knowledge Base)
and KBNL (Knowledge-Based Natural Language) and Machine Translation. Please
feel free to forward it within Digital. To obtain copies of MCC technical
reports mentioned in the Update, send mail to RDVAX::MCC (include mail stop).
Please see note regarding the distribution at the end of report.
KBNL and Machine Translation
============================
New KBNL Release
----------------
A new version of KBNL software has been released. The DEC platform version
uses Lucid Lisp on the DECstation. Major features, in comparison to the
summer release are:
o A generalized query facility has been introduced into the system -
There have been versions of this before (e.g., the old search-for command,
the Scan query interface). What is new is a functional interface to this
facility so that it can easily be embedded into whatever applications one
wants. In this facility, queries can be input in NL (either English or
Spanish). They are then understood. If there is ambiguity, various
disambiguation strategies are invoked. including, for example, preferring an
interpretation with a larger number of resolved anaphors, or preferring one in
which an object is known to satisfy all the subcategorizations to one which is
merely consistent with them. If these are not enough, then it tries to
paraphrase the alternatives to the user and asks which was intended. The
paraphraser uses special generation strategies that exploit specialized
knowledge about the kinds of ambiguities that can occur. These strategies
enable it to generate alternatives that make clear what the ambiguity is that
must be resolved. Once an interpretation is selected, it is translated into
the KB Query Language. That language is part of the Generic KB Interface,
aimed at portability and interoperability among KBs (see the recent tech
report ACT-NL-393-91 "A Functional Interface to a Knowledge Base: an NLP
Perspective" by Barnett et al.). Results from execution of the query are then
displayed to the user, in a mixture of textual description and graphical
display.
o The new grammatical system has been completely installed -
Over the last several months, much effort has been invested in making the
grammar representation system in KBNL easier to understand and to modify. The
syntactic processing components of both Lucy and Koko were substantially
modified in order to incorporate the new changes. Other modifications to the
syntactic system speed it up and make its operation easier to follow.
A new technique was incorporated into the system for implementing context tests,
which restrict the application of syntactic rules to those cases in which the
necessary neighbor edges are also on the chart. A tech report describes the
new system and makes it possible for users to modify the grammar if necessary
to support particular applications: ACT-NL-380-91-P "An Overview of the KBNL
Grammar Formalism" by Barnett et al.
o The Luke lexical acquisition tool is now part of the system -
The new version of Luke is independent of not only a particular KB but also a
particular knowledge representation (KR) system. This posed a bit of a problem
initially since Luke, unlike the rest of KBNL, actually needs to exploit
specific KR operations within itself. For example, it needs a simple truth
maintenance system that it can use to keep track of dependencies so that the
lexicon can be maintained as the KB changes. The new version of Luke exploits a
very simple KR system that has been implemented as part of Luke. This means
that the KR operations that Luke needs can run, regardless of the KR system
that is used for the domain KB that Luke is creating dictionary entries for.
Just as in the old system, Luke's output is a set of dictionary entries that
describe meanings in terms of the domain KB that it is given. The entries are
written to a file, which acts as a permanent dictionary. The file is then
compiled by a separate dictionary compiler to build the runtime dictionaries
that Lucy and Koko need.
Japanese Work Begun
-------------------
Work has begun on a Japanese morphological analyzer as part of the
DARPA-funded TIPSTER project. Japanese text can be read in, displayed,
and printed out. A simple segmentation system was built that can be applied
to the TIPSTER texts. It is likely that this system will be abandoned in
favor of the system being imported from the Univ. of Kyoto, but building one
helped to learn a lot about the way text is encoded in Japanese. They are now
able to look Japanese words from an input text up in the EDR Dictionary word
list and find the concept ID for the word. As a result of all of this, a
first serious step has been taken toward extending KBNL to Japanese. To do
so, they will still have to build all the major components, just as in
any new language. But the fact that Japanese uses a different orthographic
system from is no longer an issue.
CYC
===
The 1991 version of the Cyc software (Cyc-141) has been released. It runs
on DS3100 and DS5000, among other platforms.
The index to CYC topics and publications has been revised: ACT-CYC-357-91
"Annotated Index for CYC Project, Version 1.2"
The alpha version of the new, portable knowledge editing facility, ZUE,
is described in this recent tech report: ACT-CYC-296-91-Q "The ZUE Portable
Interface for CYC (Documentation for Version 0.8)"
'Grand Logic Lecture' - Tape and Report Available
-------------------------------------------------
A new videotape and technical report on Cyc and knowledge representation has
been released. The 2-hour lecture, "Talking to CYC: An Introduction to the
Theory of Knowledge Representation (Video)" (MCC TR No. ACT-CYC-343-91-Q),
covers the basic syntax and semantics of the constraint language used in
communicating with CYC, including sentence logic and basic predicate calculus.
The 94 page technical report ("Talking to CYC: An Introduction to the Theory of
Knowledge Representation", MCC technical report no. ACT-CYC-342-91) covers all
of these issues in more depth, includes an extensive introduction to set theory
and relation theory, and covers certain advanced issues such as access levels,
truth-maintenance, and the representation of modal sentences. Exercises (and
answers) are included at the end of the technical report. The video is classed
confidential and proprietary.
Cyc-West Office
---------------
There are now four Cyc researchers (i.e. "Cyclists") in the Palo Alto,
California, office. This will give CYC participants on the west coast easier
access to direct assistance and demonstrations. With remote server
communication working, operations performed in Austin are transmitted to
CYC-West and operations performed at CYC-West are transmitted to Austin.
This is likely to be the prototype mode of intercommunication between
CYC-Austin and participant company collaborator sites. To that end, they
are developing the mechanisms to keep various proprietary micro-theories
from being universally broadcast, yet checked continually at MCC for
incompatibility with all new operations and changes to the CYC KB.
If any Digital engineers have applications or wish to explore applications of
CYC, you are invited you to contact AUSTIN::KUHLMAN to discuss them. This may
include visiting either of the CYC locations for intensive training.
Knowledge entry
---------------
In the knowledge entry area, work was done on: representing several aspects
of human relationships, including the relations between coworkers, bosses and
subordinates; dimensions of social status (American culture) spelling out
"codes of conduct" (expected behaviors) for different kinds of workers and
professionals; anatomical capabilities, like arm and leg movements; weather
theories; a theory of scents; fluid flow; and a general theory of what people
want (power, good heath, enough to eat, lack of pain, etc.).
Self tests
----------
There are now a half-dozen or so self-tests which CYC is able to perform on
itself, automatically. Once the test machine is set up, it will be devoted to
running and rerunning such self-diagnoses, over and over again, 24 hours per
day. That should help catch various classes of subtle inconsistencies sooner
after they are introduced into the KB.
Visual Metaphors
================
Between 1984 and 1987, the Human Interface Program at MCC investigated
3D user interfaces and graphic realism to support knowledge workers.
A technical report reviews these efforts, which provide the foundations
for the current Mirage project: ACT-348-91 "Visual Metaphors: 3D Interfaces
for Knowledge Workers" by Poltrock et al.
============================================================================
Note on distribution:
A new set of distribution lists covering 13 MCC technology areas is being
prepared by the DEC Technology Transfer Program Office in the Mill. This
newsletter will be segmented into the AI-CYC and KBNL lists when they are
completed. Then the current distribution will be discarded. You should have
received an announcement and sign up list in early October. Please respond
to RDVAX::MCC if you wish to continue to receive mailings.
The Microelectronics and Computer Technology Corporation (MCC) is a
cooperative R&D enterprise whose mission is to strengthen and sustain the
competitiveness of member companies who share common elements of a
technical vision in information technology. MCC's membership currently
includes 22 shareholders and 49 associate members.
|
| 310.21 | FWD: MCC AI Update | MR4DEC::ROACH | TANSTAAFL ! | Thu Jan 02 1992 21:31 | 118 |
|
I N T E R O F F I C E M E M O R A N D U M
Doc. No: 007137
Date: 02-Jan-1992 01:21pm EST
From: BEANE
BEANE@SHLACT@SELECT@MRGATE@NRGATE@NRO
Dept:
Tel No:
TO: PATRICK ROACH@LMO
Subject: FWD: MCC AI Update
From: BIGRED::AUSTIN::KUHLMAN "Bill Kuhlman (512) 338-3243 31-Dec-1991 1539"
31-DEC-1991 15:49:47.39
To: @MCC-AI
CC:
Subj: MCC AI Update
MCC AI UPDATE No. 21 INTERNAL USE ONLY
********************
December 31, 1991 (95 lines)
Best Wishes for a Great New Year!
AI Update covers MCC's CYC project (Large Scale Common-Sense Knowledge Base)
and KBNL (Knowledge-Based Natural Language) and Machine Translation. Please
feel free to forward it within Digital. To obtain copies of MCC technical
reports mentioned in the Update, send mail to RDVAX::MCC (include mail stop).
KBNL and Machine Translation
============================
Japanese Text Processing
-----------------------
A Japanese morphological analyzer was obtained from Prof. Nagao at the
University of Kyoto. Performance tests have determined that it is far too
slow for the large-scale text processing task in the DARPA-funded TIPSTER
project. The major bottleneck appears to be in dictionary access. The
group is working to speed it up by caching common entries so that substantially
fewer file accesses are required.
An almost complete Japanese text processing environment is now up and running
at MCC. This turns out to be difficult, even though most of the necessary
software is public domain. A set of instructions on how to install a Japanese
text environment in Unix is expected to be available in January. Send mail
if you would like to receive a copy of it.
Semantic Redesign
-----------------
A large-scale redesign of the semantic processing component of KBNL has begun.
The goal is to make the system more powerful in general so that fewer special
cases need to be considered. Natural language semantics is, in general, an
open- ended problem. We can never hope for a complete solution, but we can
build one that will solve specific problems. Finding a good design is even
more difficult for MCC than for some other groups because of the need to
guarantee that the operations that are performed are reversible (in the sense
that they can be run in one direction for understanding and in the other
direction for generation). Both the design and the implementation of the new
system will proceed in stages. If any of you have particular needs that you
would like to have considered early on, please let us know.
CYC
===
Technical Reports
-----------------
Four new technical reports were released by CYC Project:
ACT-CYC-407-91 "Ideas for Applying CYC" by Lenat & Guha. This report lists
10 ideas for applying CYC which would be appropriate to begin work on now, for
completion in a one to five year time frame. You are encouraged to read this
report.
ACT-CYC-423-91 "Contexts: A Formalization and Some Applications" by R.V. Guha.
This is Guha's completed Ph.D. dissertation at Stanford. It is the definitive
reference on contexts (microtheories) in CYC.
ACT-CYC-434-91 "Quantities in CYC" by K. Goolsbey. This paper describes
the approach taken by the CYC project in representing and reasoning with
quantities.
ACT-CYC-406-91 "Comparing CYC to other AI Systems" by Guha & Lenat. This
report compares CYC to nine (seemingly) related AI projects that are currently
underway. The purpose is to answer questions of the sort: "I heard that
project x will be available within a few years, so why do I need Cyc?"
Knowledge Entry
---------------
Work in the knowledge entry area focused on: social status, theories of
human relationships, the code of conduct in an office setting, skill capability,
spatial relationships that exist been parts of human hands and the objects
they interact with, and things that cause pain.
AI Related OODS tech Report
===========================
ACT-OODS-433-91 "On the Semantics of Protocols Among Distributed Intelligent
Agents" by M. Singh.
Comings & Goings
================
Phil Apley and Bob Kearns, from Digital's Cambridge Research Lab, visited
the CYC project in December.
=============================================================================
The Microelectronics and Computer Technology Corporation (MCC) is a
cooperative R&D enterprise whose mission is to strengthen and sustain the
competitiveness of member companies who share common elements of a
technical vision in information technology. MCC's membership currently
includes 22 shareholders and 49 associate members.
|
| 310.22 | FWD: f. y. i | MR4DEC::ROACH | TANSTAAFL ! | Fri Jan 24 1992 15:48 | 462 |
|
I N T E R O F F I C E M E M O R A N D U M
Doc. No: 007372
Date: 21-Jan-1992 03:46pm EST
From: PAPAGEORGE
PAPAGEORGE@LMOADM@AIDEV@MRGATE@NRGATE@NRO
Dept:
Tel No:
TO: PAT ROACH@LMO
Subject: FWD: f. y. i
From: SELECT::BESNIA "JOAN BESNIA, LMO2/D3, 296 5030 21-Jan-1992 0925"
21-JAN-1992 09:31:51.09
To: kbs_group1.dis
CC: BESNIA
Subj: f. y. i
From: SELECT::JOSYULA "Be there..Be good..Be done." 20-JAN-1992
19:27:01.69
To: POLITAKIS,KELLY,BIRNBACH
CC: BESNIA
Subj: Dear Joan: Can you distribute to all DKAS?
From: ESIS::AUSTIN::KUHLMAN "Bill Kuhlman (512) 338-3243 20-Jan-1992 1504"
20-JAN-1992 16:03:39.92
To: ESIS::JOSYULA
CC:
Subj: MCC fuzzy systems project
Hello Ram,
MCC is launching a new project in the area of Fuzzy Systems. I've attached
the prospectus on the project (about 400 lines). I would appreciate it if
you
would look the prospectus over and get back to me regarding the following:
Should Digital participate in the project? Are you or your organization
interested in participating? Who else in the company should hear about this?
Also, Steve O'hara, the MCC manager of the project, will be in the Boston area
January 20 & 24, and can talk with you and/or your group about the details of
"Craft". So, please let me know if you would like to meet with Steve.
Regards,
Bill Kuhlman - MCC liaison, CRA
CRAFT PROJECT PROSPECTUS
Common Runtime Architecture for Fuzzy Technology
version 1.1, January 9, 1992
Executive Summary:
MCC is launching a new project in the area of Fuzzy Systems.
While Japan has fielded thousands of Fuzzy applications, there has
been limited activity in the U.S. It has become clear that the
two primary roadblocks to the utilization of the technology by the
broad market are the isolation of existing Fuzzy tools from
mainstream computer systems and the lack of industry (or de facto)
standards.
The focus of the CRAFT project is on bringing Fuzzy technology
into main-stream computing. In addition, the CRAFT project will
assist in the development of international standards for Fuzzy
Systems and in the imple-mentation of prototype software that
conforms to those standards. This work will be based on and
reinforced by customer-supplied applications.
Problem Addressed:
Fuzzy Logic is an emerging technology -- applications are just
beginning to appear in the U.S. There are a few commercial
vendors, but they are all small and their products do not work
well together. If a company is interested in Fuzzy Systems, they
essentially have to make a gamble on a particular vendor's tools
without any guarantee that the developed system will be compatible
with any other vendor's tools. The lack of de facto standards in
this area is inhibiting many organizations from investing in Fuzzy
Systems.
In Japan this has not been a problem because many large Japanese
companies are completely vertically integrated with respect to
computing technology with almost no dependence on external
suppliers, other than "keiretsu" style agreements. In the U.S.,
most companies have a complex web of technology suppliers which,
without standards in place, forces them to select when they would
prefer to be supplier-independent.
From the vendor's perspective, each of them is faced with the
problems of interfacing their Fuzzy products with other systems
and software. Without a standard framework for integrating Fuzzy
techniques into general purpose applications, each vendor is
forced to build their own custom solution. Once CRAFT is in
place, these problems can be solved once and the results will be
available to every vendor and end-user.
Business Objective:
The primary purpose of the CRAFT project is to develop the
necessary software for Fuzzy systems to become part of mainstream
computing. This involves two main goals.
The first goal is to provide tools for applications developers to
free them from concern about which software tool is being used and
which hardware platform(s) are needed. The CRAFT project will
develop a software architecture to provide interoperability of
Fuzzy vendors' hardware and software. Using this layer of
interoperability, applications developed for one platform will be
portable to other platforms, without modifications. As a
consequence of this development, standards will be proposed to the
appropriate International Standards bodies.
The second goal of CRAFT is to extend the software and standards
to handle real world situations where there are distributed,
heterogeneous processors running hybrid applications. In many of
today's applications, analysts are forced to select a single model
of execution and cast their problem in its terms, when the problem
could more readily be solved by using a combination of
technologies -- such as combining neural nets with Fuzzy systems
for adaptability, and interfacing with relational databases for
information access.
In addition to the above objectives, another key component to
getting companies started with Fuzzy technology is the lack of an
educational tool geared for engineers to help them assess the
suitability of the technology for their particular applications.
We expect to work with Professor Zadeh's Berkeley Initiative in
Soft Computing (BISC) to supply such a Primer. It will be
released to project participants as soon as possible and will be
made available to the general public about six months later. It
will address issues such as:
o What is Fuzzy Logic?
o What Applications has it been used for?
o How can I assess its suitability for my application?
Technical Plan:
The CRAFT project will be based on three core premises. First,
international standards will foster Fuzzy systems development both
in the U.S. and world-wide. Second, the standards must be
supported by software that implements the key features of the
standards. Third, the whole process must be validated by working
on relevant applications. An appropriate balance between these
objectives will be maintained throughout the project.
During the development of the standards, prototype code will be
developed that enables the software to be isolated from the
hardware to the extent that applications can be readily ported to
new environments without modi-fications. This prototype code will
evolve into a more general prototype that can handle the full
range of heterogeneous hardware and software platforms, including
versions for UNIX, MS-DOS, Windows, etc.
After this phase has been sufficiently resolved, a second effort
will focus on enabling applications to run in concert with other
programming styles, on mixed hardware platforms, and across
multiple platforms at the same time.
In order for this project to be successful, we have already
identified a number of critical efforts:
o Definition, design and development of a canonical execution
format, with prototype develop-ment of associated translators and
interpreters.
o Development of translation tools to accommodate existing
applications.
o Design and implementation of a runtime kernel that can be
ported to existing hardware.
o Development of mechanisms for interprocess / interprocessor
communications and synchronization.
o Establishment of a framework for cascading and / or layering
Fuzzy and non-Fuzzy routines.
o Definition and implementation of a system to allow
interoperability of heterogeneous platforms.
o Creation of an efficient framework for development of
integrated Fuzzy / Neural applications.
o Development of standard interfaces to existing DBMS and file
systems.
Perhaps the most challenging topic from a theoretical point of
view is to develop tools for assuring stability and robustness of
Fuzzy applications. The current approaches are inadequate for
many critical applications. One approach under consideration is
the development of a Fuzzy-based model of a system, which in turn
can be analyzed rigorously for stability. We intend to
investigate this area in detail during the CRAFT project, most
likely by funding university research.
Application Areas:
Typical Fuzzy applications today are rule-based systems, often in
the area of Controls. The CRAFT architecture will be designed to
accommodate as many different kinds of Fuzzy applications as
possible.
One significant area of interest is in Image Processing. There
are a number of applications of Fuzzy being researched that
involve Image Compression, Image Enhancement, Image Understanding,
etc. We will be working on applications in these areas.
Another application area is Diagnostic Expert Systems. There are
a number of systems in use today that measure numbers against
thresholds looking for possible problems in some plant or process.
By using Fuzzy Logic, the thresholds can be "smoothed" out into
continuous functions which will yield a much more accurate and
reliable diagnostic system.
The actual set of applications that we collaborate on will be
determined by the project participants -- our motivation is to
ensure broad coverage of potential Fuzzy applications.
Project Organization:
This will be a distinctly international project. Foreign
companies will be able to participate in the project definition
and development and will have the same access to results as U.S.
companies.
Although the driving force behind this project will be end-users,
who have real problems to solve, we expect participation from each
of these following groups.
<Insert graphic here showing CRAFT working with many other types
of organizations, such as universities, governments, other
research organizations, etc.>
The sponsoring organization for the CRAFT project is a newly-
formed not-for-profit subsidiary of MCC, called ATLAS, the
Advanced Technology Laboratory for the Advancement of Standards.
CRAFT personnel will manage the development of the standards and
the software development process. It is expected that major
portions of the actual work will be done by project participants,
including some work by MCC's Neural Networks Group. One of the
keys to the success of the project is the assignment of visiting
scientists to the project by member companies.
The CRAFT Directorate will be composed of members from
participating companies. This group will be responsible for
making sure that the CRAFT architecture meets the business needs
of each participant. Due to the number of companies expected to
join this project, we will probably have to offer seats on the
Directorate on a rotational basis. Because of their early
involvement, Charter Members are guaranteed a seat on the
Directorate.
There will also be a Technical Steering Committee for the project,
composed primarily of the leading (international) academic
researchers in the field. Professor Zadeh has graciously agreed
to direct this steering committee which will provide frequent
guidance to the project.
CRAFT is an international project, with participants from around
the world contributing to its development, and being able to
exploit the prototype code developed from this effort. There will
be a large number of organizations contributing to this project,
each with different resources and different motives for joining.
It is quite likely that major components of research will be
subcontracted out to third parties or universities. Likewise,
technology may be licensed from other vendors if appropriate.
Organizations that wish to provide technology or conduct research
for this project are encouraged to contact us.
Project Staffing:
Due to the nature of the problem addressed, a range of resources
will be used to develop the CRAFT software and standards. When
companies commit to participating the project, they can also
commit some dedicated personnel resources. This approach has the
dual benefit of "jump-starting" the project and ensures that
resulting technology will be transferred back to the funding
organizations.
We envision the following project personnel requirements:
Sub-project Optimal Size Initial Size
Core Run-time Architecture 4-5 2-3
Hybrid Systems Development 2-3 1-2
Distributed & Heterogeneous Dev. 5-6 2-3
Integration and Testing 3-4 1-2
Technology Transfer 2-3 -
Management and Administration 4 2
Total number of people 20-25 8-12
At least one fourth of the people are expected to be CRAFT direct
hire employees while the remainder are expected to be Visiting
Scientists from participant companies. [The Visiting Scientist
policy provides for business travel expenses, office space,
equipment and supplies, but does not pay for base salary,
benefits, living expenses or relocation.]
If significant geographic pockets of activity are identified and
if sub-projects can be developed relatively independently of the
other sub-projects, then regional offices may be justified. We
would, however, like to encourage cooperation and interchange
between project participants.
Technology Transfer Plan:
1. Design and prototype the CRAFT software to support
interoperability and general purpose interfaces.
2. Develop a series of customer-specific applications using
various vendor's software and hardware to test the CRAFT software.
3. Use these applications as a basis for developing a proposed
standard for interfacing various types of software with various
types of hardware. This interface layer, named CRAFT (Common Run-
time Architecture for Fuzzy Technology), will be published and
promoted as an international standard via such Standards bodies as
ISO and IEEE.
4. Refine and generalize the prototype code for delivery to all
project participants.
Timeline:
Jan-Feb '92 Charter Member sign-up period.
Jan-Dec '92 Sign-up period, with incentives for early enrollment.
Apr 6, '92 Project launch.
Nov 1, '92 Fuzzy Primer completed and distributed to
project participants.
Jun 30, '93 Interoperability standards, alpha version complete.
Aug 31, '94 Hybrid and Distributed/Heterogeneous capabilities.
Dec 31, '94 Stability final results published for
participants.
Mar 31, '95 Final deliverables, Project conclusion. Final
documentation released.
Apr 1, '95 Start of extended support period.
DEC contact:
Bill Kuhlman, MCC liaison
AUSTIN::KUHLMAN
512-338-3243
MCC contact:
Steven O'Hara, Acting Manager
MCC CRAFT Fuzzy Systems Project
3500 W. Balcones Center Drive
Austin, Texas 78759 U.S.A.
(512) 338-3776
fax: (512) 338-3885
e-mail: [email protected]
CRAFT PROJECT STRUCTURE SUMMARY
Common Runtime Architecture for Fuzzy Technology
Apr 6, 1992 to Mar 31, 1995 (3 years)
version 1.0, January 9, 1992
Base Fee
These fees are valid through March 31, 1992. After that, the fees increase as
described on the next page. This fee is based on a full three year
commitment. Foreign participants are asked to pay a small premium over U.S.
and Canadian companies in order to cover the larger costs of international
support and travel.
N. America $125k per year, for three years.
Foreign $150k per year, for three years.
Prototype software developed will remain the property of MCC, will be licensed
for commercial use to all participants, and will be supported for at least six
months from the date of project termination. An unlimited internal use
license is included. Default royalty payments will be 3% of gross revenues
from resale of derivative works, with a lifetime cap of $3 million, from
products and services that directly utilize the CRAFT software. A royalty-
free option is detailed below.
Charter Membership
A limited number of Charter Memberships will be offered until February 28,
1992. These members will receive a 20 percent discount for the first year and
will be guaranteed a permanent seat on the CRAFT Directorate. Other seats on
the Directorate will be offered on an annual rotational basis to non-Charter
Members.
Visiting Scientist Option
Participants are encouraged to consider providing one full time Visiting
Scientist. A limited number of Visiting Scientists will be accepted.
Qualifications for these scientists will be determined by CRAFT project
management. Base salary, relocation and benefits will be provided by the
sponsoring company while business travel expenses, office space, equipment and
supplies will be provided by the CRAFT project. Under these terms, the fee is
reduced:
N. America $75k per year, plus one Visiting Scientist.
Foreign $95k per year, plus one Visiting Scientist
Option for No Royalty Payments
Under this option companies will receive a royalty-free, perpetual worldwide
license to commercially exploit the CRAFT software.
This option can be considered prepaying royalties on the following basis: 10%
of the $3 million cap is $300k; which comes to $100k per year for three years.
Visiting Fee per year Fee per year
Country Scientist? no royalties with royalties
N. America No $225k $125k
N. America Yes $175k $75k
Foreign No $250k $150k
Foreign Yes $195k $95k
Small Vendors of Fuzzy Products and Services
For small companies that cannot afford the base fee, special arrangements will
be made that involve lower payments and higher royalties.
Late Entry Fees (after March 31, 1992)
The final opportunity to join at the above prices is March 31, 1992. After
that point, all prices increase by 3% per month and are still based on three
full years of payments, compressed down to the remaining term. No additional
participants in the CRAFT project will be accepted after December 31, 1992.
For example, a U.S. company with a Visiting Scientist could pay $6,250 a month
if they joined during March 1992. If they joined in September 1992 (six
months late), the monthly payments for the remaining 30 months would be
$8,960.
|
| 310.23 | Fwd: December, 1991 MCC Technical Report Abstracts | MR4DEC::ROACH | TANSTAAFL ! | Fri Jan 24 1992 16:10 | 114 |
|
I N T E R O F F I C E M E M O R A N D U M
Doc. No: 007383
Date: 22-Jan-1992 05:23pm EST
From: BEANE
BEANE@SHLACT@SELECT@MRGATE@NRGATE@NRO
Dept:
Tel No:
TO: PATRICK ROACH@LMO
Subject: Fwd: December, 1991 MCC Technical Report Abstracts & Revisio
THIS IS FOR DIGITAL INTERNAL USE ONLY
Title: Recognizing Overlapping Hand-Printed Characters by
Centered-Object Integrated Segmentation and Recognition
Author(s): Gale L. Martin
Number: ACT-NN-395-91
Date: October, 1991
Visual object recognition is often conceived of as a final step in a visual
processing system. First, physical information in the raw image is used to
isolate and enhance to-be-recognized clumps and then each of the resulting
pre-processed representations is fed into the recognizer. This general
conception fails when there are no reliable physical cues for isolating the
objects, such as when objects overlap. This paper describes an approach,
called centered object integrated segmentation and recognition (COISR), for
integrating object segmentation and recognition within a single neural
network. The application is hand-printed character recognition. The approach
uses a back-propagation network that scans a field of characters and is trained
to recognize whether it is centered over a single character or between
characters. When it is centered over a character, the net classifies the
character. The approach is tested on a dataset of hand-oriented digits and
high accuracy rates are reported.
********************************************************************************
Title: Fuzzy Control Systems
Author(s): Eric Hartman, Philip D. Heermann
Number: ACT-NN-402-91
Date: November, 1991
Fuzzy control systems began to appear in the mid-1980's in large numbers of
commercial and consumer products. Japanese companies have led the way with
the development of fuzzy control chips and applications/products ranging from
air-conditioners, washing machines and video cameras to railway systems,
highway tunnel ventilation systems and seaport cranes, as well as stock market
trading systems and character recognition systems [8]. In this paper we
compare and contrast fuzzy control to other methods and discuss some common
criticisms.
The is For Digital Internal Use Only
December, 1991 MCC Technical Report Abstracts & Revisions
Title: Discovery as Autonomous Learning from the Environment
Author(s): Wei-Min Shen
Number: ACT-OODS-387-91
Date: October, 1991
It seems plausible to hypothesize that discovery involves collaboration among
many intelligent activities. However, there is little knowledge about how and
in what form such collaborations occur. In this paper, a framework is
proposed to view discovery as learning from the environment by autonomous
systems. Within this framework, many intelligent activities such as
perception, action, exploration, experimentation, learning, problem solving,
and new term construction can be integrated in a coherent way. The framework
is presented in detail through an implemented system called LIVE. To evaluate
this approach, performance of LIVE in several discovery tasks are also shown.
The conclusion is that autonomous learning from the environment is a vital
approach for deepening our understanding about discovery in general.
********************************************************************************
Title: On the Semantics of Protocols Among Distributed Intelligent
Agents
Author(s): Munindar P. Singh
Number: ACT-OODS-433-91
Date: December, 1991
The continuing expansion of distributed intelligent systems makes new demands
on theories of communication in Computer Science. It is customary to
describe the individual nodes or agents in an intelligent system in
terms of higher level concepts like intentions, know-how and beliefs.
However, current theories of the communication among such agents provide
no form of a formal or rigorous semantics for the messages exchanged at a
corresponding level of abstraction - they either concern themselves
with implemenational details or address what is, for artificial systems, an
irrelevant aspect of the problem. A recent theory of communication that gives
the objective model-theoretic semantics for speech acts is applied to
this problem. This allows the important properties of protocols to be
formalized abstractly, i.e., at the level of the application, not of the
implementation. Further constraints on "good" designs can also be stated,
which simplify the requirements imposed on the member agents. The resulting
theory not only provides some insights into designing distributed intelligent
systems, but also helps in their validation. As an example, it is applied
to a logical reconstruction of the classical Contract Net protocol.
THIS IS FOR DIGITAL INTERNAL USE ONLY
Title: Technical Issues of Enterprise Integration
Author(s): Alexander Cavalli, Roy Smith
Number: EID-349-91
Date: November 1991
Computing infrastructure will be a part of Enterprise Integration solutions.
This report summarizes a set of technical issues that must be addressed in
developing Enterprise Integration systems.
|