Conference heron::euro_swas_ai

306.0. "FWD: MCC AI Update" by ULYSSE::ROACH (TANSTAAFL !) Tue Mar 26 1991 12:04

 

                  I N T E R O F F I C E   M E M O R A N D U M

                                        Date:     05-Mar-1991 00:16am CET
                                        From:     BEANE
                                                  BEANE@BIGRED@MRGATE@DPD20@DPD
                                        Dept:      
                                        Tel No:    

TO: See Below

Subject: FWD: MCC AI Update

From:	AUSTIN::KUHLMAN "Bill Kuhlman (512) 338-3243  27-Feb-1991 1217"   27-FEB-1991 
12:21:29.32
To:	@MCC-AI
CC:	
Subj:	MCC AI Update

MCC AI UPDATE No. 13                                     INTERNAL USE ONLY
********************
February 27, 1991                                              (156 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
=== 
New Software Release
--------------------
A new version (140) of CYC software was released on February 15.  The first 
edition of the FI (Functional Interface) Manual, a handbook for CYC application 
builders, was included as one of the documents accompanying the release.  Also 
included with the release are users' manuals for both the new KE Region 
knowledge entering tool, and the UE (Unit Editor) on Symbolics machines:

  ACT-CYC-063-91-Q  "CYC 140/14 KB 67 (Software Tapes 1 & 2)"
  ACT-CYC-064-91-Q  "CYC 140 Documentation (One Tape)"
 
  ACT-CYC-021-91-Q  "FI Manual" by D. Pratt
 
  ACT-CYC-018-91-Q  "CYC KE-Region User's Guide (for CYC version 140)"
 
  ACT-CYC-065-91-Q  "CYC User's Manual for Interface on Symbolics Machines 
  (for CYC version 140)"

The fourth in a series of technical reports was completed which describes 
the principles behind and contents of the CYC KB.  It deals with an assortment
of issues ranging from the CYCL language to representation details:

  ACT-CYC-022-91  "The World According to CYC, Part 4" by Lenat et al.

Porting
------- 
The port of CYC and UE to the DECstation 3100 mentioned last month has begun. 
As a part of this effort, core CYC (CYCL, the KB, etc.) was ported to the 3100 
under Lucid Common Lisp.  Phill Apley, from Digital's Cambridge Research Lab, 
and Bob Kearns came to Austin for a week to participate in the specification 
for a new CYC interface to be written in CLIM.  Bob Kerns is presently working 
as a consultant to CYC on-site at DEC CRL along with Phill Apley to implement 
the first version of that interface.  

The Macintosh II version of CYC was brought up to the current Symbolics version.
Some bugs were fixed in the Macintosh II CYC interface and the Macintosh 
display-oriented CYC editor.  The KE-Region feature was ported to the Mac using 
MACL's version of the EMACS editor.  This currently works reasonably well 
running locally, but still requires some work to run under the remote 
functional interface.

A set of portable editor functions was written for CYC and a new subsystem 
for utilities that work with the host environments version of EMACS.  

Knowledge Entry
--------------- 
The main efforts on the knowledge entering front were concentrated on entering 
knowledge about building-parts, general buildings, and related script-types.  
There was much wrestling with fundamental issues of relationships between 
microtheories.  Work progressed on NaivePhysics and HumanAilment microtheories.
 
KBNL (Knowledge-Based Natural Language) and Machine Translation
=============================================================== 
Work focused on two areas.  The first is translation between languages where 
lexical distinctions do not line up. For example, English "fish" goes to 
Spanish "pez" or "pescado", depending on whether the fish has been caught.  
English "river" goes to French "fleuve" or "riviere" depending on whether 
or not it flows to the ocean.  The simplest available technique for dealing 
with this problem just chooses one form and uses it all the time, thus often 
making mistakes.  MCC's early attempts to solve this  problem failed when 
there were other, irrelevant distinctions also present in the lexicon (e.g., 
for fish, between trout, salmon, etc.). Other techniques required a 
cross-language lexical comparison, which  is expensive even though it can be 
done off-line and cached.  The current method avoids both of these problems by 
requiring that lexical items have one additional piece of information 
associated with them, namely marked/unmarked in particular respects.  Then, if 
a direct mapping produces an unmarked form, (e.g., "fish" ->"pez"), it knows to 
check the applicability conditions of the marked forms along that dimension 
(e.g., if caught then "pescado").  This information is specific to one language 
(rather than to a language pair) and it can be acquired semiautomatically with 
Luke.
 
The second thrust of the work is Luke, which is being rewritten to be knowledge 
base independent.  The result of this effort will be a version of Luke that 
can be combined with any knowledge base that provides substantial assistance 
for lexical acquisition.  One benefit of this effort to better organize Luke's 
knowledge is that the task of porting Luke to new languages will be much more 
straightforward.

Natural Language Tech Reports
-----------------------------
ACT-NL-066-91  "Planning, Reacting, and Communicating"  by E. Rich
  This paper presents an outline of a model of language use, including both 
  generation and understanding.  The model is a hybrid that exploits both 
  planning and reaction.
 
ACT-NL-019-91  "A Functional Interface to a Knowledge Base for Use by a 
  Natural Language Processing System" by Barnett et al.  The KBNL system
  assumes that it has access to a knowledge base that describes both general 
  concepts in the world as well as concepts in the domain(s) that are being
  considered.  This report describes the functional interface that KBNL
  assumes exists between itself and the knowledge base.

ACT-NL-003-91  "Capturing Language-Specific Semantic Distinctions in Interlingua
  Based MT"  by Barnett et al.  This paper describes an interlingua-based 
  approach to machine translation motivated by two goals:  maximizing the
  ability of the system to produce correct translations in cases of "non-
  information preserving translation", and minimizing the cost of adding a
  new language to the system.

Other AI Tech Reports
---------------------
ACT-364-90  "Context Maintenance" by C. Petrie.  Traditional applications
  of "Truth Maintenance Systems" (TMSs) fail to adequately represent heuristic
  search in which some paths are initially preferred.  This paper show how
  a problem solver/TMS architecture called REDUX maintains the validity of
  the reason for context choices and rejections.

ACT-AI-014-91  "DETENTE: Practical Support for Practical Action"  by Wroblewski
  et al.  This paper presents an object-oriented system to embed personal
  agendas in complex application interfaces.

ACT-AI-048-91  "Person-Computer Cooperation Through Collaborative Manipulation"
  by L. Terveen.  This report contains the text of a UT dissertation done in 
  the MCC Human Interface Lab.  An approach to the design of cooperative
  computer systems is embodied in the HITS Knowledge Editor (HKE).

ACT-AI-052-91  "The Construction of Human-Computer Interfaces Considered as
  as Craft" by D. Wroblewski.  This paper advances the notion that building
  human interfaces is a craft, and examines the implications for management,
  research, and teaching.

================================================================================
Monthly 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:  SHIRLEY CRIDER@DVO
TO:  STEVE DONOVAN@DLO
TO:  DENNIS DICKERSON@DLO
TO:  Gale Kleinberger@HSO
TO:  Mike Sievers@HSO
TO:  Mike Willis@HSO
TO:  Sherry Williams@HSO
TO:  Katherine Jones@HSO
TO:  Dale Stout@HSO
TO:  Tommy Gaut@HSO
TO:  Tom Wilson@HST
TO:  jim rather@HSO