nl-kr-request@cs.rpi.edu (NL-KR Moderator Chris Welty) (04/20/89)
NL-KR Digest (Thu Apr 20 11:49:56 1989) Volume 6 No. 22 Today's Topics: Parsing Hungarian: Answer to Nurkkala IJCAI-89 Workshop on Conceptual Graphs BBN AI Seminar: Paul Cohen BBN AI Seminar: Gunar Liepins CSLI Calendar, April 20, 4:23 Submissions: nl-kr@cs.rpi.edu Requests, policy: nl-kr-request@cs.rpi.edu Back issues are available from host archive.cs.rpi.edu [128.213.1.10] 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. ----------------------------------------------------------------- To: nl-kr@cs.rpi.edu Date: Tue, 18 Apr 89 08:19:10 +0200 >From: Klaus Schubert <dlt1!schubert@nluug.nl> Phone: +31 30 911911 Telex: 40342 bso nl Subject: Parsing Hungarian: Answer to Nurkkala An answer to Tom Nurkkala: You are looking for references to work on languages with an affix-based syntax. According to Mel'^cuk's judgement ("English is very exotic") any language other than English will do. I cannot offer Finnish, which would match your name best, but Hungarian. The article is written by a scholar who has written a Hungarian word parser and deals mainly with the underlying morphological analysis, suited for dependency-syntactic parsing: Pr'osz'eky, G'abor (1988): Hungarian - a special challenge to machine translation? In: New Directions in Machine Translation. Ed. Dan Maxwell / Klaus Schubert / Toon Witkam. Dordrecht / Providence: Foris Publications, pp. 219-231 Regards, Klaus Schubert ------------------------------ To: nl-kr@cs.rpi.edu Date: 19 Apr 89 09:22:13 EDT >From: john Sowa <SOWA@ibm.com> Subject: IJCAI-89 Workshop on Conceptual Graphs Call for Participation: Workshop on Conceptual Graphs August 20 & 21, 1989 IJCAI-89 The Fourth Annual Workshop on Conceptual Graphs will be held at IJCAI-89 in Detroit, Michigan, on Sunday August 20 and Monday August 21. It will provide a forum for researchers and practitioners to exchange ideas about the theory and applications of conceptual graphs. Attendance will be limited to people who are actively using, developing, extending, or implementing conceptual graphs. Those who are interested in participating should submit a two-page extended abstract about their work with an indication of whether they would like to (a) present a full paper, (b) present a short summary of their work, or (c) simply attend. Seven copies of the abstract are due by May 10 at the following address: Conceptual Graph Workshop Committee c/o Janice A. Nagle 1641 E. Old Shakopee Road Bloomington, MN 55425 Copies of the proceedings of the 1988 Conceptual Graph Workshop are available from the AAAI for $20. No proceedings are available for the first two workshops. ------------------------------ To: nl-kr@cs.rpi.edu >From: Marc Vilain <mvilain@BBN.COM> Subject: BBN AI Seminar: Paul Cohen Date: Wed, 19 Apr 89 14:30:37 EDT BBN STC Science Development Program AI Seminar Series Lecture PLAUSIBLE INFERENCE, EXTENDED COMPOSITION, AND ONTOLOGY MAINTENANCE PAUL R. COHEN Experimental Knowledge Systems Laboratory Department of Computer and Information Science University of Massachusetts, Amherst BBN STC, 2nd floor large conference room 10 Moulton St, Cambridge MA, 02138 Friday April 21st, 10:30 AM I will present work I have done with Cynthia Loiselle on a simple method for generating rules of plausible inference from the relations in a knowledge base, and, more recently, on the question of how to predict the plausibility of the conclusions of inferences. Unlike deductive inferences, conclusions generated by rules of plausible inference are not *guaranteed* to be "true" or plausible in any sense, so for every rule, we need to know whether it generates plausible conclusions (or, in the case of Collins' certainty conditions, what would make the conclusions more or less plausible). Experiments with human subjects show that relatively little information is needed to make moderately accurate plausibility predictions for the rules we generated. Still, roughly 30% of the implausible inferences in our test set were predicted to be plausible, so we have been examining what additional knowledge is necessary to improve this performance. I will describe Huhns and Stephens' adaptation of relation element theory (called extended composition) to the task of predicting the plausibility of inferences, and show that their system is essentially equivalent to our own, but could be extended to provide the information needed to improve plausibility predictions. I will also touch on the role of plausible inference in ontology maintenance, the process of determining the meaning of new relations or revising the meaning of existing relations in a very large knowledge base such as CYC. ------------------------------ To: nl-kr@cs.rpi.edu >From: Marc Vilain <mvilain@BBN.COM> Date: Wed, 19 Apr 89 15:35:55 EDT Subject: BBN AI Seminar: Gunar Liepins BBN STC Science Development Program AI Seminar Series Lecture ISSUES IN GENETIC OPTIMIZATION GUNAR LIEPINS Oak Ridge National Laboratories BBN STC, 2nd floor large conference room 10 Moulton St, Cambridge MA, 02138 Tuesday April 25th, 10:30 AM This presentation reviews several genetic algorithm applications, provides a brief introduction to the genetic paradigm, and addresses multi-objective and constrained optimization. The roles of sampling (embedding) and representation are made explicit and illustrated from the perspective of function dimensionality and smoothness. The four modes of GA failure: estimation, crossover disruption, stability of regions of attraction, and schemata deceptiveness, are reviewed. A simple construction for fully deceptive problems of arbitrary size is given. The presentation concludes that the primary challenges are to improve GA efficiency and better characterize their domain of applicability. ------------------------------ To: nl-kr@cs.rpi.edu Date: Wed, 19 Apr 89 16:33:52 PDT >From: emma@csli.Stanford.EDU (Emma Pease) Subject: CSLI Calendar, April 20, 4:23 C S L I C A L E N D A R O F P U B L I C E V E N T S _____________________________________________________________________________ 20 April 1989 Stanford Vol. 4, No. 23 _____________________________________________________________________________ A weekly publication of The Center for the Study of Language and Information, Ventura Hall, Stanford University, Stanford, CA 94305 ____________ CSLI ACTIVITIES FOR THIS THURSDAY, 20 April 1989 2:15 p.m. CSLI Seminar Cordura Hall Varieties of Context: Session 3 Conference Room Indexicality in Context Geoffrey Nunberg, Xerox PARC (nunberg.pa@xerox.com) Respondent: Brian Smith 3:30 p.m. Tea Ventura Hall 4:00 p.m. STASS Seminar Cordura Hall Dewey on Defeasible Reasoning Conference Room Tom Burke (burke@csli.stanford.edu) Abstract in last week's Calendar ____________ CSLI ACTIVITIES FOR NEXT THURSDAY, 27 April 1989 12:00 p.m. TINLunch Cordura Hall Reading: "A compositional approach to discourse Conference Room representation theory" by Henk Zeevat Discussion led by Stanley Peters (peters@csli.stanford.edu) Abstract below 2:15 p.m. CSLI Seminar Cordura Hall Varieties of Context: Session 4 Conference Room Contexts in Activity Lucy Suchman, Xerox PARC (suchman.pa@xerox.com) Respondent: Susan Stucky Abstract below 3:30 p.m. Tea Ventura Hall 4:00 p.m. STASS Seminar Cordura Hall To be announced Conference Room ____________ NEXT WEEK'S TINLUNCH Reading: "A compositional approach to discourse representation theory" by Henk Zeevat (Linguistics and Philosophy 12:95-131, 1989) Discussion led by Stanley Peters (peters@csli.stanford.edu) April 27 Hans Kamp and others have talked about the system for representing discourse, which he introduced in "A theory of truth and semantic representation," as both a formal language for analyzing the contribution sentences make to the truth conditions of a discourse and an account of the representation that hearers construct mentally of a situation or world being described to them. These authors have emphasized the noncompositionality of discourse representation systems. Zeevat's paper attempts to show how Kamp's DRT can be formulated as a compositional system within Montague's framework of Universal Grammar, and thereby to isolate what is really distinctive about DRT as contrasted with Montague's Intensional Logic and other more "traditional" ways of representing truth conditions. ____________ NEXT WEEK'S CSLI SEMINAR Varieties of Context: Session 4 Contexts in Activity Lucy Suchman, Xerox PARC (suchman.pa@xerox.com) Respondent: Susan Stucky April 27 In "Plans and Situated Actions" (1987) I argued that the central problem for an account of purposeful action is to understand the relation between our reasoning about action and the organization of our activity in situ. Such an understanding requires that we take efficient descriptions like plans as resources for rather than determinants of the organization of situated activity. In this talk I'll review the argument briefly, drawing a parallel between this view and the architecture of John Perry's three-story house. In particular, I'll locate the problem of situated activity as the role of the second floor, the missing middle, in establishing a productive interaction between intentions and the requirements of locally contingent action. Recently, we have embarked on a new project to explore the problem of situated activity in a specific setting. A starting premise for the investigation is that rather than there being a context that surrounds this setting's activity and gives it sense, participants in the setting continually reproduce multiple domains of relevance which organize their actions and to which their actions are made accountable. We'll work through this premise and some of its implications in relation to a piece of activity recorded on videotape. ____________ SYMBOLIC SYSTEMS FORUM Symbols and Thought Fred Dretske, Philosophy Friday, 28 April, 3:15, 60:62N Symbols have meaning. Hence, to manipulate symbols is to operate on---or at least with---meaningful elements. And this, according to some, is what both minds do when they think, reason, and infer and what machines (mainly digital computers) do when they multiply, renumber footnotes, or correct our spelling. Hence, a computer's manipulation of symbols is an attractive model for the mind. Both are symbol systems---systems that traffic in meanings. If thinking that the sun is shining was manipulating (in some appropriate way) a symbol, or a set of symbols, that meant that the sun is shining, then there would be reason to believe that machines could (or would some day---as soon as we got them to manipulate symbols the right way) think that the sun is shining. Is this right? Is thinking that the sun is shining merely a matter of manipulating (in the right way) symbols that mean the sun is shining? If not, what more is required? ------------------------------ End of NL-KR Digest *******************