nl-kr-request@cs.rpi.edu (NL-KR Moderator Chris Welty) (10/04/90)
NL-KR Digest (Wed Oct 3 23:46:11 1990) Volume 7 No. 17 Today's Topics: LOOM, Classifiers & ExpertSystems CFP: Processing Declarative Knowledge (LaTeX Format) THIRD ISAI IN MEXICO CFP - INHERITANCE IN NATURAL LANGUAGE PROCESSING INHERITANCE IN NATURAL LANGUAGE PROCESSING AI SEMINAR ANNOUNCEMENT: Patrick Love, Sept. 24 AI SEMINAR REMINDER: Hammond, Sept 21, 2 pm Submissions: nl-kr@cs.rpi.edu Requests, policy: nl-kr-request@cs.rpi.edu Back issues are available from host archive.cs.rpi.edu [128.213.5.17] in the files nl-kr/Vxx/Nyy (ie nl-kr/V01/N01 for V1#1), mail requests will not be promptly satisfied. If you can't reach `cs.rpi.edu' you may want to use `turing.cs.rpi.edu' instead. BITNET subscribers: we now have a LISTSERVer for nl-kr. You may send submissions to NL-KR@RPIECS and any listserv-style administrative requests to LISTSERV@RPIECS. ----------------------------------------------------------------- To: nl-kr@cs.rpi.edu >From: avesani@irst.it (Paolo Avesani) Newsgroups: comp.ai,comp.ai.nlang-know-rep Subject: LOOM, Classifiers & ExpertSystems Date: 7 Sep 90 07:49:27 GMT We are looking for real world applications developed with the LOOM system, the Knowledge Representation system produced by Robert MacGregor and John Yen at ISI Laboratory. We are interested in any documentation concerning expert systems, fully or partially developed with LOOM. Our goal is to evaluate with real experiences the capabilities of this knowledge representation system. In particular we are interested in the use of classifiers in an expert system project and about in LOOM as a developing system. Expert System Group I.R.S.T. (Istituto per la Ricerca Scientifica e Tecnologica) Please email to: Avesani@irst.uucp Avesani@irst.it Avesani%irst@uunet.uu.net (from ARPA) ------------------------------ To: nl-kr@cs.rpi.edu >From: workshop organizers <pdk@uklirb.informatik.uni-kl.de> Newsgroups: comp.ai.nlang-know-rep Subject: CFP: Processing Declarative Knowledge (LaTeX Format) Date: 28 Sep 90 09:37:08 GMT Reply-To: workshop organizers <pdk@uklirb.informatik.uni-kl.de> \documentstyle[11pt]{article} \addtolength{\evensidemargin}{-1.8cm} \addtolength{\oddsidemargin}{-1.8cm} \addtolength{\textwidth}{3.6cm} \addtolength{\topmargin}{-2.5cm} \addtolength{\textheight}{7.5cm} \unitlength1.0cm \setlength{\parindent}{0.0cm} \renewcommand{\baselinestretch}{0.9} \begin{document} \thispagestyle{empty} \vspace*{1cm} \begin{center} {\Huge\bf PDK '91} \\ \vspace{1cm} {\large CALL FOR PAPERS} \\ \vspace{1cm} {\Large\bf International Workshop} \\ \vspace{0.6cm} {\large\bf on} \\ \vspace{0.6cm} {\huge\bf Processing Declarative Knowledge} \\ \vspace{0.6cm} {\Large\bf --- Representation and Implementation Methods ---} \\ \vspace{1.2cm} {\bf July 1-3, 1991 Kaiserslautern, Germany} \vspace{1.2cm} \end{center} \setlength{\parindent}{0cm} {\bf Background:} \\ \setlength{\parindent}{1.0cm} Declarative representation formalisms have long constituted the kernels of AI languages. Their high description level facilitates readability, maintenance, and parallelization of knowledge bases; their orientation toward logic enables clear semantics. This becomes especially important when several formalisms are used in a `hybrid' fashion, as the amalgamation of sublanguages is simplified. However, the processing of large declarative knowledge bases is becoming efficient only with the use of modern implementation techniques. For instance, the increased gap to von~Neumann machines may be bridged by (global) static analysis and (multi-stage) transformation/compilation of the representation formalisms. Experimental methods for most representation and implementation layers exist, which should gradually enter practical AI, e.g. via expert-system shells. \\[0.2cm] \setlength{\parindent}{0cm} {\bf Scope:} \\ \setlength{\parindent}{1.0cm} This workshop will provide an opportunity to present promising approaches for processing declarative knowledge, to demonstrate implemented systems, and to meet with AI practitioners. Also welcome are well-founded critiques of declarative-knowledge processing (e.g. from procedural, object-oriented, or connectionist points of view). Besides representation formalisms on the basis of horn-logic programming it is possible to present, among others: more general {\it inference rules} (backward/forward chaining), {\it concept-description languages} (subsumption procedures), as well as {\it constraint} or {\it constraint-logic-program\-ming systems} (propagation algorithms). Regarding implementation, all recent interpretation and especially compilation techniques will be of interest, e.g.: {\it abstract interpretation}, {\it partial evaluation}, {\it rule compilation}, and {\it WAM technology}. \newpage \thispagestyle{empty} \setlength{\parindent}{0cm} {\bf Paper Submission:} \\ \setlength{\parindent}{1.0cm} Please submit {\bf four (4) copies} of papers written in English by {\bf 1 March 1991} (not via email). If electronic mail is available, additionally email an ASCII version of the abstract. Refereeing will take place in two categories: {\bf Long papers} of max. 20 pages for full-blown research results, {\bf short papers} of up to\ 6 pages for concise presentations and partial results. In both categories we expect original work. Authors will be {\bf notified of acceptance or rejection} of submitted papers by {\bf 30 April 1991}, the {\bf camera-ready revisions} are due by {\bf 10 June 1991}. {\bf Applications for system presentations} consisting of a system mini-description of 1-3 pages as well as a specification of the hardware/software required should arrive by {\bf 3 June 1991}. Please direct contributions and requests to the following address: \begin{tabbing} \hspace{3cm}\= {\bf \+PDK}\\ {\bf DFKI GmbH}\\ {\bf P.~O.~Box 2080}\\ {\bf 6750 Kaiserslautern, F.~R.~Germany} \\[0.15cm] FAX: +49-631-205-3210 \\ email: pdk@informatik.uni-kl.de \end{tabbing} Preprints of refereed papers will be produced for the workshop, and a proceedings publication is scheduled immediately afterwards. \\[0.2cm] \setlength{\parindent}{0cm} {\bf Associated Societies:} \\ \setlength{\parindent}{1.0cm} The workshop is organized by the German Research Center for Artificial Intelligence (DFKI) in cooperation with the Association for Logic Programming (ALP) and the Gesellschaft fr Informatik e.V. (GI). \\[0.2cm] \setlength{\parindent}{0cm} {\bf Important Dates:} \begin{tabbing} \hspace*{1cm}\=30 April 1991: \hspace{0.5cm}\= \kill \> 1 March 1991: \> Deadline for submission of papers \\ \> 30 April 1991: \> Notification of acceptance or rejection \\ \> 3 June 1991: \> Deadline for application of system demonstrations\\ \> 10 June 1991: \> Camera-ready paper\\ \> 1-3 July 1991: \> Workshop\\ \end{tabbing} {\bf Program Committee:} \\ {\small \begin{tabbing} Maurice Bruynooghe, Kath. Universiteit Leuven \hspace{0.5cm}\= \kill Hassan A\"{\i}t-Kaci, DEC Paris \> Alexander Herold, ECRC Munich \\ H.-J. Appelrath, University of Oldenburg \> Robert Kowalski, Imperial College London \\ Woody Bledsoe, University of Texas at Austin \> Hans Langmaack, University of Kiel \\ Egon B{\"o}rger, University of Pisa \> Jean-Louis Lassez, IBM Yorktown Heights/NY \\ Harold Boley, DFKI Kaiserslautern \> Michael M. Richter, DFKI Kaiserslautern (Chair) \\ Maurice Bruynooghe, Kath. Universiteit Leuven \> Erik Sandewall, Univerity of Link{\"o}ping \\ Tim Finin, Unisys Paoli \> John Taylor, Hewlett Packard Bristol \\ Herv{\'e} Gallaire, Bull Paris \> A. Voronkov, Int. Lab. of Intelligent Syst. Novosibirsk \\ Jan Grabowski, Humboldt University Berlin \end{tabbing}} \end{document} ------------------------------ To: nl-kr@cs.rpi.edu Date: Fri, 28 Sep 90 09:40:06 CST >From: "Centro de Inteligencia Artificial(ITESM)" <ISAI@TECMTYVM.MTY.ITESM.MX> Subject: THIRD ISAI IN MEXICO THIRD INTERNATIONAL SYMPOSIUM ON ARTIFICIAL INTELLIGENCE: APPLICATIONS OF ENGINEERING DESIGN & MANUFACTURING IN INDUSTRIALIZED AND DEVELOPING COUNTRIES OCTOBER 22-26, 1990 ITESM, MEXICO The Third International Symposium on Artificial Intelligence will be held in Monterrey, N.L. Mexico on October 22-26, 1990. The Symposium is sponsored by the ITESM (Instituto Tecnologico y de Estudios Superiores de Monterrey) in cooperation with the International Joint Conferences on Artificial Intelligence Inc., the American Association for Artificial Intelligence, the Sociedad Mexicana de Inteligencia Artificial and IBM of Mexico. GOALS: * Promote the development and use of AI technology in the solution of real world problems. Analyze the state-of-the-art of AI technology in different countries. Evaluate efforts made in the use of AI technology in all countries. FORMAT: ISAI consists of a tutorial and a conference. Tutorial.- Oct. 22-23 Set of seminars on relevant AI topics given in two days. Topics covered in the Tutorial include: "Expert Systems in Manufacturing" Mark Fox, Ph.D., Carnegie Mellon University, USA "A.I. as a Software Development Methodology" Randolph Goebel, Ph.D., University of Alberta, Canada Conference.- Oct. 24-26 Set of lectures given during three days. It consists of invited papers and selected papers from the "Call for Papers" invitation. Areas of application include: computer aided product design, computer aided product manufacturing, use of industrial robots, process control and ES, automatic process inspection and production planning. Confirmed guest speakers: Nick Cercone, Ph.D, Simon Fraser University, Canada "Expert Information Management with Integrated Interfaces" Mitsuru Ishizuka, Ph.D, University of Tokyo, Japan "Fast Hypothetical Reasoning System as an Advanced Knowledge-base Framework" Alan K. Mackworth, Ph.D, University of British Columbia, Canada "Model-based Computational Vision" Antonio Sanchez, Ph.D, Universidad de las Americas, Mexico Sarosh N. Talukdar, Ph.D, Carnegie Mellon University, USA "Desing System Productivity: Some bottlenecks and potential solutions". Carlos Zozaya Gorostiza, Ph.D, CONDUMEX, Mexico IMPORTANT: Computer manufacturers, AI commercial companies, universities and selected papers with working programs could present products and demonstrations during the conference. In order to encourage an atmosphere of friendship and exchange among participants, some social events are being organized. For your convinience we have arranged a free shuttle bus service between the hotel zone and the ITESM during the three day conference. FEES: Tutorial.- Professionals $ 250 USD + Tx(15%) Students $ 125 USD + Tx(15%) Conference.- Professionals $ 180 USD + Tx(15%) Students $ 90 USD + Tx(15%) Simultaneous Translation $ 7 USD Formal dinner $ 25 USD * *(Includes dinner, open bar, music (Oct 26)) Tutorial fee includes: Tutorial material. Welcoming cocktail party (Oct.22) Conference fee includes: Proceedings. Welcoming cocktail party (Oct.24) Cocktail party. (Oct.25) HOTELS: Call one to the hotels listed below and mention that you are going to the 3rd. ISAI. Published rates are single or double rooms. HOTEL PHONE* RATE Hotel Ambassador 42-20-40 $85 USD + Tx(15%) Gran Hotel Ancira 42-48-06 $75 USD + Tx(15%) 91(800) 83-060 Hotel Monterrey 43-51-(20 to 29) $60 USD + Tx(15%) Hotel Rio 44-90-40 $48 USD + Tx(15%) * The area code for Monterrey is (83). REGISTRATION PROCEDURE: Send personal check payable to "I.T.E.S.M." to: "Centro de Inteligencia Artificial, Attention: Leticia Rodriguez, Sucursal de Correos "J", C.P. 64849, Monterrey, N.L. Mexico" INFORMATION: CENTRO DE INTELIGENCIA ARTIFICIAL, ITESM. SUC. DE CORREOS "J", C.P. 64849 MONTERREY, N.L. MEXICO. TEL. (83) 58-20-00 EXT.5132 or 5143. TELEFAX (83) 58-07-71, (83) 58-89-31, NET ADDRESS: ISAI AT TECMTYVM.BITNET ISAI AT TECMTYVM.MTY.ITESM.MX ------------------------------ To: nl-kr@cs.rpi.edu >From: walter@kub.nl Date: Thu, 27 Sep 90 10:41:21 +0200 Subject: CFP - INHERITANCE IN NATURAL LANGUAGE PROCESSING X-Envelope-To: nl-kr-request@cs.rpi.edu CALL FOR PAPERS COMPUTATIONAL LINGUISTICS Special issue on INHERITANCE IN NATURAL LANGUAGE PROCESSING Editors: Walter Daelemans & Gerald Gazdar Inheritance mechanisms and their implications for description, theory and processing have become an increasingly important research topic in computational linguistics. Structure sharing by inheritance in frame-based and object-oriented knowledge representation has been exploited by computational linguists since the mid-seventies. At first, use was mainly for the representation of semantic and world knowledge, but, in the 1980s, this usage has extended to the syntactic, morphological and phonological domains. Some linguistic and computational linguistic frameworks have even explicitly incorporated inheritance into their theories. Concurrently, there have been major advances in the understanding of general theories of inheritance (especially default inheritance and the nonmonotonic reasoning that it supports) by AI researchers in the field of knowledge representation. Papers dealing with any aspect of the application of inheritance techniques within computational linguistics are solicited for the special issue. The deadline for submission of full papers is March 1, 1991. A total of seven copies of each submission is required: send four copies to one of the editors of the special issue, and three copies to the editor of Computational Linguistics. Addresses of the special issue editors: Walter Daelemans Institute for Language Technology and AI (ITK) Tilburg University P.O. Box 90153 5000 LE Tilburg, The Netherlands Tel: (13) 663070 walter@kub.nl Gerald Gazdar School of Cognitive and Computing Sciences University of Sussex Brighton BN1 9QH, UK Tel: (273) 678029 geraldg@cogs.sussex.ac.uk Address of the editor of Computational Linguistics: James F. Allen Computer Science Department University of Rochester Rochester, NY 14627, USA Tel: (716) 275-5288 acl@rochester.arpa __________________________________________________________________________ Note: in the summer of 1990 an international workshop on Inheritance in Natural Language Processing was held in Tilburg. The proceedings, with abstracts and preliminary versions of the papers presented, can be ordered from Nicole Lubeck, ITK, P.O.Box 90153, 5000 LE Tilburg, The Netherlands, +31 13 663113, lubeck@kub.nl. Enclose a cheque for 25 Dfl (30 Dfl overseas) with your order, payable to ITK. __________________________________________________________________________ ------------------------------ To: nl-kr@cs.rpi.edu >From: walter@kub.nl Date: Thu, 27 Sep 90 10:41:21 +0200 Subject: INHERITANCE IN NATURAL LANGUAGE PROCESSING X-Envelope-To: nl-kr-request@cs.rpi.edu REPORT ON THE FIRST INTERNATIONAL WORKSHOP ON INHERITANCE IN NATURAL LANGUAGE PROCESSING Wietske Sijtsma ITK, Tilburg University The use of the notion of inheritance is well-established in both object-oriented programming language theory and knowledge representation. More recently linguistic theories have emerged in which inheritance is used for the representation of various sorts of linguistic knowledge. This workshop brought together linguists, logicians and AI-researchers interested in and working with inheritance theory. Keynote speaker Rich Thomason provided a general overview of the use of inheritance theory in the field of knowledge representation. He noted that non-monotonic reasoning and multiple inheritance appear to be required when representing world knowledge. Yet both notions introduce complications: non-monotonic reasoning introduces the the problem of exceptions to a generalisation (the "Tweetie triangle") while multiple inheritance introduces the problem of conflicting information (the "Nixon diamond"). Thomason looks with hope to the use of inheritance theory in the domain of Natural Language Processing: here is a restricted domain with well-defined problems. Insights obtained from the use of inheritance theory in NLP could prove benificial to knowledge representation. He also expressed a word of caution, however, about the proliferation of formalisms. In his paper, Walter Daelemans provided the arguments in favour of the use of inheritance theory in NLP. Some of the arguments are well-established: default reasoning allows for reasoning with incomplete information, the notion of information sharing implies a minimisation of redundancy, a structured way of representing blocking effects and markedness differences, etc. Daelemans also argued that the use of the notion of multiple inheritance can in fact replace the main functions now performed by unification. He argued that the need to introduce the notion of defaults in unification-based grammar formalisms has led to sometimes akward extensions of the UBG formalisms. Daelemans claims that replacing the latter by feature-based formalisms with multiple inheritance would lead to more elegant solutions. In this context it was interesting to see the extensions Bouma proposed to Categorial Unification Grammar, the UBG formalisms he works on. Bouma has introduced the notion of default unification to CUG. With the help of default unification Bouma is able to override default information in a number of formally defined contexts. I.e., default unification introduces the possibility of non-monotonic reasoning in CUG. In Word Grammar, the formalism Fraser & Hudson work on, inheritance has always played a crucial role. In order to deal with the existence of exceptions to generalisations, the NOT proposition has been introduced in WG. The idea is that in case a general morphological rule (eg. plural formation) or syntactic transformation (eg. subject-auxiliary inversion) does not apply this is encoded by explicitly denying the application of the generalisation in question. It must be observed that, as one of the participant's of the workshop remarked, Fraser & Hudson's proposal implies that more exceptional cases are at times represented as being less exceptional than actual less exceptional cases (using the term exceptional in the sense of actually occurring). Both Bouma and Fraser & Hudson explicitly deny the need to introduce the notion of multiple inheritance, the mechanism Daelemans claims to be a suitable replacement of unification. Bouma, while recognising the need for multiple inheritance in representing world knowledge, doubted whether pure linguistic data exist which actually call for the introduction of multiple inheritance. The outcome of the discussion that followed was that in those cases where linguists have introduced multiple inheritance, technical alternatives are available. A choice between the various mechanisms available should take such matters as notational adequacy into account. The usefulness of the notion inheritance in NLP was perhaps first recognised by lexicologists and morphologists. As Russell, Carroll & Warwick noted, lexical organisation requires a means to express generalisations on the one hand and exceptions on the other, means which are supplied by inheritance theory and default reasoning. Not surprisingly then, a number of papers revolved around the use of inheritance mechanism in building lexicons. Ann Copestake discussed the AQUILEX project, which aims at extracting taxonomies from machine readable dictionaries. The idea is to build lexical knowledge bases with the help of the taxonomies extracted. The taxonomies are in fact very similar to inheritance hierarchies. Various problems arise while extracting taxonomies from machine readable dictionaries. Copestake is well aware of these problems; eg. the use of cycles in dictionary definitions, the need for sense disambiguation when building a taxonomy. Instead of trying to solve the various problems that occur, the aim of the AQUILEX project is to derive further taxonomies and to classify the problems encountered before defining possible solutions. De Smedt & de Graaf presented a model of Dutch morphology in which structured inheritance plays a crucial role. Structured inheritance implies that information is not simply copied from one object in a hierarchy to the next, but rather that a "prototypical" object is defined on which all associated objects are modeled. The obvious advantage of structured inheritance is the massive reduction of redundancy in the specification of a network. The paper presented by Flickinger and Nerbonne also falls in the lexicalist tradition. Working within the framework of Head-driven Phrase Structure Grammar, they showed how certain syntactic varations can be accounted for in the lexicon. They define an inheritance hierarchy of word classes in which both syntactic, semantic and subcategorisation information is stored. By allowing the explicit overriding of lexical information, they are able to capture important generalisations over both word classes and the distribution of certain properties in the lexicon. In his talk, Thomason already foreshadowed a future of inheritance theory in which formalisation would play an increasingly important role. Bob Carpenter in fact provided a glimpse of what such a more formalised future might entail. He proposed various new and borrowed extensions to the theory of feature structures. The main extension involves the typing of feature structures and the requirement that every type is associated with a specific set of appropriate features and feature values. The types are organised in a subsumption hierarchy in which for each type appropriate features and possible values are declared. Unification is then defined for all well-typed feature structures. The workshop showed that inheritance is a mechanism employed in many areas of NLP. Some areas in which inheritance proves useful have already been discussed. But also in the field of pragmatics (Mercer) and semantics (Morreau) such notions as default reasoning prove extremely useful. Considering the numerous areas in which inheritance theory is used, the work of Evans & Gazdar on the complexity of DATR, a declarative language for representing inheritance networks, is of particular importance. If linguists, logicians and AI-researchers find more and more uses for inheritance theory it is obvious that knowledge on the formal properties becomes of crucial importance. The workshop provided an extensive and interesting overview of the areas in which inheritance is used. In the Proceedings of the workshop the organizers refer to the ever growing interest in the use of inheritance in NLP. Let's hope that the interest for the topic will result in a Second International Workshop. ------------------------------ To: nl-kr@cs.rpi.edu Date: Thu, 20 Sep 90 09:39 EDT >From: MMETEER@rcca.bbn.com Subject: AI SEMINAR ANNOUNCEMENT: Patrick Love, Sept. 24 [ Note: these talks are in the past, included for people who like to know what they've missed - CW ] BBN Science Development Program AI Seminar Series Lecture THE USE AND ABUSE OF KNOWLEDGE-BASED TECHNOLOGY WHILE MAKING ALUMINUM ARTIFACTS DR. PATRICK LOVE ALCOA Technical Center ALCOA Center, PA 15069 love@alcoa.com BBN, 2nd floor large conference room 10 Moulton St, Cambridge MA, 02138 10:30 am, Monday, September 24, 1990 ALCOA has had an AI group as part of the corporation since 1983. It has grown to a size of some 13 people although it seems inaccurate to call it an AI group today. In this talk I will give a brief review of some of the projects that the group has worked on since 83, how those projects are viewed today, how the group has changed, and what lessons can be learned from our corporate experience with knowledge-based technology. Two interesting issues intersect the technical lessons. The first is the current wave in industry to sort out all the vendors and products and develop some internal standards from what computer managers view as industry chaos. The second is the large technical mismatch that often occurs between MIS managers who have the responsibility to keep systems running and LISP machine wizards who want to reboot production systems when the window-manager gets munged. ******************************************************* Suggestions for AI Seminar speakers are always welcome. Please e-mail suggestions to Marie Meteer (mmeteer@bbn.com) or Dan Cerys (cerys@bbn.com). ******************************************************* ------------------------------ To: nl-kr@cs.rpi.edu Date: Thu, 20 Sep 90 15:00 EDT >From: MMETEER@rcca.bbn.com Subject: AI SEMINAR REMINDER: Hammond, Sept 21, 2 pm ***************************** * REMINDER * ***************************** BBN Science Development Program AI Seminar Series Lecture PLANNING, MEMORY AND OPPORTUNISM: A CASE-BASED APPROACH TO THE CONTROL OF ACTION Kristian J. Hammond The University of Chicago Artificial Intelligence Laboratory 1100 East 58th Street Chicago, IL 60637 (kris@gargoyle.uchicago.edu) BBN, 2nd Floor Large Conference Room 10 Moulton St., Cambridge, MA 2:00 p.m., Friday, September 21, 1990 Over the past few years, the thrust of research in planning has moved away from the notion of a plan as a proof and more towards an approach that integrates planning and execution. Rather than building systems that produce plans, researchers are more concerned with systems that produce behavior in the face of a changing world. This change in focus has opened up a wide variety of new problems in the area of planning and activity. In this talk, I will discuss one view of some new issues in planning and outline an approach for dealing with them that combines case-based planning with a reactive component and a learning mechanism that uses run-time expectation failures to focus the planner's attention on gaps in its knowledge base. In terms of planning, the system is able to recover from many execution-time failures as well as capitalize on unforeseen opportunities. In terms of learning, it is able to learn from both run-time opportunities and unexpected execution problems. I will also introduce the Chicago model of planning as part of a theory of autonomous agency that includes planning, action, learning and the long-term control of the agent's environment. ------------------------------ End of NL-KR Digest *******************