nl-kr-request@CS.ROCHESTER.EDU (NL-KR Moderator Brad Miller) (11/03/87)
NL-KR Digest (11/02/87 18:28:58) Volume 3 Number 42 Today's Topics: Seminars: SUNY Buffalo Cog. Sci. Colloq.: Contini-Morava BBN AI Seminar Reminder -- Amy Lansky Learning in Connectionist Networks BBN AI Seminar -- Jeffrey Siskind Speech Recognition Using Connectionist Networks (UNISYS) From CSLI Calendar, Oct. 29, 3:5 Linguistics in Encyclopedia of AI ---------------------------------------------------------------------- Date: Fri, 23 Oct 87 14:51 EDT From: William J. Rapaport <rapaport@cs.Buffalo.EDU> Subject: SUNY Buffalo Cog. Sci. Colloq.: Contini-Morava STATE UNIVERSITY OF NEW YORK AT BUFFALO GRADUATE GROUP IN COGNITIVE SCIENCE ELLEN CONTINI-MORAVA Department of Linguistics University of Virginia TEMPORAL EXPLICITNESS IN EVENT CONTINUITY IN SWAHILI DISCOURSE Swahili verb sequences consisting of an inflected form of `kuwa', "to be", followed by another inflected verb are constructions in which `kuwa' supplies deictic orientation for verbs whose orientation is not obvious from context. Such explicit orientation is needed in situations where there is a break in continuity between events, such as introduc- tion of a new subject, that might cause difficulty in integrating a verb into its context. The argument is supported by examples from texts and quantitative data. It is suggested that the notion of event continuity in discourse involves more than purely temporal relatedness between events. Friday, October 30, 1987 3:30 P.M. Baldy 684, Amherst Campus Informal discussion at 8:00 P.M. on Friday evening at David Zubin's house, 157 Highland Ave., Buffalo. Call Bill Rapaport (Dept. of Com- puter Science, 636-3193 or 3181) or Gail Bruder (Dept. of Psychology, 636-3676) for further information. ------------------------------ Date: Fri, 23 Oct 87 16:43 EDT From: Marc Vilain <MVILAIN@G.BBN.COM> Subject: BBN AI Seminar Reminder -- Amy Lansky BBN Science Development Program AI Seminar Series Lecture LOCALIZED EVENT-BASED REASONING FOR MULTIAGENT DOMAINS Amy L. Lansky Artificial Intelligence Center, SRI International (LANSKY@VENICE.AI.SRI.COM) BBN Labs 10 Moulton Street 2nd floor large conference room 10:30 am, Monday October 26 This talk will present the GEM concurrency model and GEMPLAN, a multiagent planner based on this model. Unlike standard state-based AI representations, GEM is unique in its explicit emphasis on events and domain structure -- a world domain is modeled as a set of regions composed of interrelated events. Event-based temporal logic constraints are then associated with each region to delimit legal domain behavior. GEM's emphasis on constraints is directly reflected in the architecture of the GEMPLAN planner -- it can be viewed as a general purpose constraint satisfaction facility. Its task is to construct a network of interrelated events that satisfies all applicable regional constraints and also achieves some stated goal. A key focus of our work has been on the use of localized techniques for domain representation and reasoning. Such techniques partition domain descriptions and reasoning tasks according to the regions of activity within a domain. For example, GEM localizes the applicability of domain constraints and also imposes additional ``locality constraints'' based on domain structure. Together, constraint localization and locality constraints help solve several aspects of the frame problem for multiagent domains. The GEMPLAN planner also reflects the use of locality; its constraint satisfaction search space is subdivided into regional planning search spaces. By explicitly utilizing constraint localization, GEMPLAN can pinpoint and rectify interactions among regional search spaces, thereby reducing the burden of ``interaction analysis'' ubiquitous to most planning systems. ------------------------------ Date: Sun, 25 Oct 87 17:34 EST From: Michael Friendly <FRIENDLY%YORKVM1.BITNET@wiscvm.wisc.edu> Subject: Seminar - Learning in Connectionist Networks Cognitive Science Discussion Group Speaker : Geoffrey Hinton, University of Toronto Title : "Learning in connectionist networks" Date : Friday, Oct. 30, 1pm Location: Rm 207 Behavioural Science Bldg., York University 4700 Keele St., Downsview, Ont. Abstract Parallel networks of simple processing elements can be trained to compute a function by being shown examples of input and output vectors. The network stores its knowledge about the function as the strengths of interactions between pairs of processors. For networks with many layers of processors between the input and output, the learning procedure must decide which aspects of the input vector need to be represented by the internal processors. By choosing to represent important underlying features of the task domain, the network can learn to model the function efficiently in the strengths of the interactions between processors, and it will then generalize appropriately when presented with new input vectors. These parallel networks are much more similar to the brain than conventional computers and may provide insight into the basis of natural intelligence. ------------------------------ Date: Tue, 27 Oct 87 10:38 EST From: Marc Vilain <MVILAIN@G.BBN.COM> Subject: BBN AI Seminar -- Jeffrey Siskind BBN Science Development Program AI Seminar Series Lecture DEPENDENCY DIRECTED PROLOG Jeffrey Mark Siskind MIT Laboratory for Computer Science (also: summer intern at Xerox PARC) (Qobi@ZERMATT.LCS.MIT.EDU) BBN Labs 10 Moulton Street 2nd floor large conference room 10:30 am, Tuesday November 3 In this talk I will describe an implementation of pure Prolog which uses dependency directed backtracking as a control strategy for pruning the search space. The implementation uses a strategy whereby the Prolog program is compiled into a finite set of templates which characterize a potentially infinite boolean expression which is satisfiable iff there is a proof of the goal query. These templates are incrementally unraveled into a sequence of propositional CNF SAT problems and represented in a TMS which is used to find solutions using dependency directed backtracking. The technique can be extended to use ATMS-like strategies for searching for multiple solutions simultaneously. Two different strategies have been implemented for dealing with unification. The first compiles the unification constraints into SAT clauses and integrates them in the TMS along with the and/or goal tree produced by unraveling the templates. The second uses a separate module for doing unification at run time. This unifier is novel in that it records dependencies and allows nonchronological retraction. The interface protocol between the TMS and the unifier module has been generalized to allow integration of other "domains" of predicates, such as linear arithmetic and simple linear inequalities, to be built into the system while still preserving the soundness and completeness of the pure logical interpretation of Prolog. In the talk, time permitting, I will discuss the search prunning advantages of this approach and its relation to previous approaches, the implementation mechanism, and some recent work indicating the potential applicability of this approach to parsing with disjunctive feature structures, such as done with the LFG and related grammar formalisms. ------------------------------ Date: Tue, 27 Oct 87 15:37 EST From: Tim Finin <finin@bigburd.PRC.Unisys.COM> Subject: Speech Recognition Using Connectionist Networks (UNISYS) AI Seminar UNISYS Knowledge Systems Paoli Research Center Paoli PA SPEECH RECOGNITION USING CONNECTIONIST NETWORKS Raymond Watrous Siemens Corporate Research and University of Pennsylvania The thesis of this research is that connectionist networks are adequate models for the problem of acoustic phonetic speech recognition by computer. Adequacy is defined as suitably high recognition performance on a representative set of speech recognition problems. Six acoustic phonetic problems are selected and discussed in relation to a physiological theory of phonetics. It is argued that the selected tasks are sufficiently representative and difficult to constitute a reasonable test of adequacy. A connectionist network is a fine-grained parallel distributed processing configuration, in which simple processing elements are interconnected by simple links. A connectionist network model for speech recognition has been defined called the TEMPORAL FLOW MODEL. The model incorporates link propagation delay and internal feedback to express temporal relationships. It has been shown that temporal flow models can be 'trained' to perform successfully some speech recognition tasks. A method of 'learning' using techniques of numerical nonlinear optimization has been demonstrated for the minimal pair "no/go", and voiced stop consonant discrimination in the context of various vowels. Methods for extending these results to new problems are discussed. 10:00am Wednesday, November 4, 1987 Cafeteria Conference Room Unisys Paloi Research Center Route 252 and Central Ave. Paoli PA 19311 -- non-UNISYS visitors who are interested in attending should -- -- send email to finin@prc.unisys.com or call 215-648-7446 -- ------------------------------ Date: Wed, 28 Oct 87 19:39 EST From: Emma Pease <Emma@CSLI.Stanford.EDU> Subject: From CSLI Calendar, Oct. 29, 3:5 [Excerpted from CSLI Calendar] Reading: "Cognitive Significance and the New Theories of Reference" by John Perry Discussion led by Bob Moore October 29 In this paper, John Perry replies to Howard Wettstein's article "Has Semantics Rested on a Mistake?" Wettstein has argued that the New Theory of Reference (actually a family of theories based on the notion of direct reference) cannot handle puzzles posed by Frege concerning the cognitive significance of language. Since Wettstein finds the arguments for the New Theory absolutely convincing, he is driven to the conclusion that semantics has nothing to say about cognitive significance. Perry argues that this is an overly pessimistic assessment, and that Frege's puzzles can be solved by drawing a distinction between the conditions under which an utterance expresses a true proposition and the proposition expressed. Perry's principal claim is that, while the New Theorists have mainly concerned themselves with the latter, it is the former that should be identified with cognitive significance. Thus arguments designed to show that the proposition expressed by an utterance cannot be its cognitive significance are irrelevant, and a broader theory of semantics can and should account for both. In the discussion, I would like to raise the issue of whether getting the semantics of propositional attitude reports right forces a tighter connection between cognitive significance of an utterance and the proposition expressed by an utterance than either Wettstein or Perry is prepared to allow for. An Introduction to Situated Automata Part II: Applications Stan Rosenschein November 5 This is the second of two lectures on the situated-automata approach to the analysis and design of embedded systems. This approach seeks to ground our understanding of embedded systems in a rigorous, objective analysis of their informational properties, where information is modeled mathematically in terms of correlations between states of the system and conditions in the environment. In the first talk we motivated the general framework, presented the central mathematical ideas on how information is carried in the states of automata, and related the mathematical properties of the model to key theoretical issues in AI, including the nature of knowledge, its representation in machines, the role of syntactic deduction, "nonmonotonic" reasoning, and the relation of knowledge and action. Some general technological implications of the approach, including reduced reliance on conventional symbolic inference and increased opportunities for parallelism were discussed. In the second lecture, I will describe the application of the situated-automata perspective to specific problems arising in the design of integrated intelligent agents, including problems of perception, planning and action selection, and linguistic communication. ------------------------------ Date: Fri, 30 Oct 87 15:05 EST From: William J. Rapaport <rapaport@cs.Buffalo.EDU> Subject: Linguistics in Encyclopedia of AI I created the following for use in my natural-language understanding course, and thought some other people might find it useful. It may be a slightly idiosyncratic list: In order for it not to become a complete list of _all_ articles in the _Encyclopedia_, I did not include articles on knowledge representation or some others on topics that readers might have thought were relevant. Please let me know if there are any outright errors. William J. Rapaport Assistant Professor Dept. of Computer Science, SUNY Buffalo, Buffalo, NY 14260 (716) 636-3193, 3181 uucp: ..!{ames,boulder,decvax,rutgers}!sunybcs!rapaport internet: rapaport@cs.buffalo.edu [if that fails, try: rapaport%cs.buffalo.edu@relay.cs.net or: rapaport@buffalo.csnet ] bitnet: rapaport@sunybcs.bitnet ======================================================================== A GUIDE TO LINGUISTICS ARTICLES IN The Encyclopedia of Artificial Intelligence Stuart C. Shapiro (editor) (John Wiley & Sons, 1987). compiled by William J. Rapaport Department of Computer Science SUNY Buffalo Buffalo, NY 14260 rapaport@cs.buffalo.edu Volume 1: (1) Hull, J. J., "Character Recognition," pp. 82-88. (2) Ballard, B., & Jones, M., "Computational Linguistics," pp. 133-151. (3) Hardt, S., "Conceptual Dependency," pp. 194-199. (4) Hindle, D., "Deep Structure," pp. 230-231. (5) Scha, R.; Bruce, B. C.; & Polanyi, L., "Discourse Understanding," pp. 233-245. (6) Woods, W. A., "Grammar, Augmented Transition Network," pp. 323-333. (7) Bruce, B., & Moser, M. G., "Grammar, Case," pp. 333-339. (8) Coelho, H., "Grammar, Definite Clause," pp. 339-342. (9) Gazdar, G., "Grammar, Generalized Phrase Structure," pp. 342-344. (10) Joshi, A., "Grammar, Phrase Structure," pp. 344-351. (11) Burton, R., "Grammar, Semantic," pp. 351-353. (12) Berwick, R., "Grammar, Transformational," p. 353-361. (13) Mallery, J. C.; Hurwitz, R.; & Duffy, G., "Hermeneutics," pp. 362-376. (14) Hill, J. C., "Language Acquisition," pp. 443-452. (15) Newmeyer, F. J., "Linguistics, Competence and Performance," pp. 503- 508. (16) Wilks, Y., "Machine Translation," pp. 564-571. (17) Tennant, H., "Menu-Based Natural Language," pp. 594-597. (18) Koskenniemi, K., "Morphology," pp. 619-620. (19) McDonald, D. D., "Natural-Language Generation," pp. 642-655. (20) Bates, M., "Natural-Language Interfaces," pp. 655-660. (21) Carbonell, J. G., & Hayes, P. J., "Natural-Language Understanding," pp. 660-677. Volume 2: (22) Petrick, S., "Parsing," pp. 687-696. (23) Riesbeck, C. K., "Parsing, Expectation-Driven," pp. 696-701. (24) Small, S. L., "Parsing, Word-Expert," pp. 701-708. (25) Keyser, S. J., "Phonemes," pp. 744-746. (26) Webber, B., "Question Answering," pp. 814-822. (27) Dyer, M.; Cullingford, R.; & Alvarado, S., "Scripts," pp. 980-994. (28) Smith, B. C., "Self-Reference," pp. 1005-1010. (29) Sowa, J., "Semantic Networks," pp. 1011-1024. (30) Hirst, G., "Semantics," pp. 1024-1029. (31) Woods, W., "Semantics, Procedural," pp. 1029-1031. (32) Allen, J. F., "Speech Acts," pp. 1062-1065. (33) Allen, J., "Speech Recognition," pp. 1065-1070. (34) Allen, J., "Speech Synthesis," pp. 1070-1076. (35) Briscoe, E. J., "Speech Understanding," pp. 1076-1083. (36) Lehnert, W. G., "Story Analysis," pp. 1090-1099. END OF FILE ------------------------------ End of NL-KR Digest *******************