AIList-REQUEST@SRI-AI.ARPA (AIList Moderator Kenneth Laws) (10/21/85)
AIList Digest Sunday, 20 Oct 1985 Volume 3 : Issue 150 Today's Topics: Seminars - Program Logics (UPenn) & Meaning, Information and Possibility (UCB) & Machine Learning and Knowledge Representation (NU) & Intelligent Mail Manipulation (MIT) & A Logic for Defeasible Rules (Buffalo) & Learning From Multiple Analogies (GTE) & Computational Discourse Analysis Using DEREDEC (MIT) & RESEARCHER and Patent Analogies (CMU) ---------------------------------------------------------------------- Date: Mon, 14 Oct 85 19:26 EDT From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA> Subject: Seminar - Program Logics (UPenn) REASONING ABOUT PROGRAMS: CONCEPTUAL AND METHOLOGICAL DISTINCTIONS DANIEL LEIVANT, COMPUTER SCIENCE, CMU 3:00 pm Tuesday, October 15, 1985 216 Moore, University of Pennsylvania Reasoning about programs can be done explicitly, in first-order or higher-order mathematical theories, or implicitly, in modal logics of programs (Hoare Logic, Dynamic Logic...). One wants the latter, but the former are better suited for metamathematical analysis (semantics, calibration of proof-theoretic strength). However, modal logics are interpretable in explicit theories, so we can eat the cake and have it. In particular, we can distinguish in modal logics of programs a purely logical component and an analytical component. For example, Hoare's Logic captures exactly logical reasoning about partial-correctness assertions over WHILE-programs. We argue that this type of completeness is more informative than relative completeness. ------------------------------ Date: Wed, 16 Oct 85 14:22:50 PDT From: admin@ucbcogsci.Berkeley.EDU (Cognitive Science Program) Subject: Seminar - Meaning, Information and Possibility (UCB) BERKELEY COGNITIVE SCIENCE PROGRAM Cognitive Science Seminar - IDS 237A Tuesday, October 22, 11:00 - 12:30 240 Bechtel Engineering Center Discussion: 12:30 - 1:30 in 200 Building T-4 ``Meaning, Information and Possibility'' L. A. Zadeh Computer Science Division, U.C. Berkeley Our approach to the connection between meaning and information is in the spirit of the Carnap--Bar-Hillel theory of state descriptions. However, our point of departure is the assump- tion that any proposition, p, may be expressed as a generalized assignment statement of the form X isr C, where X is a variable which is usually implicit in p, C is an elastic constraint on the values which X can take in a universe of discourse U, and the suffix r in the copula isr is a variable whose values define the role of C in relation to X. The principal roles are those in which r is d, in which case C is a disjunctive con- straint; and r is c, p and g, in which cases C is conjunctive, probabilistic and granular, respectively. In the case of a disjunctive constraint, isd is written for short as is, and C plays the role of a graded possibility distribution which asso- ciates with each point (or, equivalently, state-description) the degree to which it can be assigned as a value to X. This possibility distribution, then, is interpreted as the informa- tion conveyed by p. Based on this interpretation, we can con- struct a set of rules of inference which allow the possibility distribution of a conclusion to be deduced from the possibility distributions of the premises. In general, the process of inference reduces to the solution of a nonlinear program. The connection between the solution of a nonlinear program and the traditional methods of deduction in first-order logic are explained and illustrated by examples. ELSEWHERE ON CAMPUS William Clancy of Stanford University will speak on ``Heuristic Classification'' at the SESAME Colloquium on Monday, Oct. 21, 4:00pm, 2515 Tolman Hall. Ruth Maki of North Dakota State University will speak on ``Meta- comprehension: Knowing that you understand'' at the Cognitive Psychology Colloquium, Friday, October 25, 4:00pm, Beach Room, 3105 Tolman Hall. ------------------------------ Date: Thu, 17 Oct 85 15:02 EDT From: Carole D Hafner <HAFNER%northeastern.csnet@CSNET-RELAY.ARPA> Subject: Seminar - Machine Learning and Knowledge Representation (NU) Northeastern University College of Computer Science Colloquium 4p.m. Wednesday, October 30 Brittleness, Tunnel Vision, Machine Learning and Knowledge Representation Prof. Steve Gallant Northeastern University A system is brittle if it fails when presented with slight deviations from expected input. This is a major problem with knowledge representation schemes and particularly with expert systems which use them. This talk defines the notion of Tunnel Vision and shows it to be a major cause of brittleness. As a consequence it will be claimed that commonly used schemes for machine learning and knowledge representation are pre- disposed toward brittle behavior. These include decision trees, frames, and disjunctive normal form expressions. Some systems which are free from tunnel vision will be described. Place: 405 Robinson Hall Northeastern University 360 Huntington Ave. Boston MA ------------------------------ Date: Sun, 13 Oct 1985 16:53 EDT From: Peter de Jong <DEJONG%MIT-OZ at MIT-MC.ARPA> Reply-to: Cog-Sci-Request%MIT-OZ Subject: Seminar - Intelligent Mail Manipulation (MIT) [Forwarded from the MIT bboard by SASW@MIT-MC.] Thursday 17, October 4:00pm Room: NE43- 8th floor Playroom The Artificial Intelligence Lab Revolving Seminar Series "The Information Lens: An Intelligent System for Finding, Filtering, and Sorting Electronic Messages" Thomas W. Malone MIT Sloan School of Management This talk will describe an intelligent system to help people share, filter, and sort information communicated by computer-based messaging systems. The system exploits concepts from artificial intelligence such as frames, production rules, and inheritance networks, but it avoids the unsolved problems of natural language understanding by providing users with a rich set of semi-structured message templates. A consistent set of "direct manipulation" editors simplifies the use of the system by individuals, and an incremental enhancement path simplifies the adoption of the system by groups. The talk will also include an overview of the other projects and research goals in the Organizational Systems Laboratory at MIT. ------------------------------ Date: Fri, 18 Oct 85 08:30:03 EDT From: "William J. Rapaport" <rapaport%buffalo.csnet@CSNET-RELAY.ARPA> Subject: Seminar - A Logic for Defeasible Rules (Buffalo) UNIVERSITY AT BUFFALO STATE UNIVERSITY OF NEW YORK DEPARTMENT OF COMPUTER SCIENCE COLLOQUIUM DONALD NUTE Advanced Computational Methods Center and Department of Philosophy University of Georgia A LOGIC FOR DEFEASIBLE RULES Humans reason using defeasible and sometimes conflicting rules like `Matches burn when struck' and `Wet things don't burn'. A formal language for representing sentential versions of such rules is presented together with a derivability relation for this language. The resulting system, LDR, is non-monotonic. Inspired by work in conditional logic, the non-monotonic rules of LDR correspond to simple subjunctive and `might' conditionals. Chaining of these rules is restricted in LDR just as the transi- tivity of the conditional is restricted in conditional logics. Several notions of consistency and coherency are defined. LDR is of special importance for research in automated reasoning, since its language is PROLOG-like and its derivability relation can be implemented in PROLOG. Thursday, November 7, 1985 3:30 P.M. Bell 337, Amherst Campus Wine and cheese will be served at 4:30 P.M., 224 Bell Hall For further information, contact: William J. Rapaport Assistant Professor Dept. of Computer Science, SUNY Buffalo, Buffalo, NY 14260 (716) 636-3193, 3181 uucp: ...{allegra,decvax,watmath}!sunybcs!rapaport ...{cmc12,hao,harpo}!seismo!rochester!rocksvax!sunybcs!rapaport cs/arpanet: rapaport%buffalo@csnet-relay ------------------------------ Date: Fri, 18 Oct 85 23:44:39 EDT From: Bernard Silver <SILVER@MIT-MC.ARPA> Subject: Seminar - Learning From Multiple Analogies (GTE) GTE LABS INCORPORATED MACHINE LEARNING SEMINAR Title: Learning from Multiple Analogies Speaker: Mark H. Burstein BBN Labs. Date: Monday October 21, 10am Place: GTE Labs 40 Sylvan Rd, Waltham MA 02254 Students learning about an unfamiliar new subject under the guidance of a teacher or textbook, are often taught basic concepts by analogies to things that they are more familiar with. Although this seems to be a very powerful form of instruction, the process by which students make use of this kind of instruction has been little studied by AI learning theorists. A cognitive process model of how students make use of such analogies will be presented. The model was motivated by examples of the behavior of several students who were tutored on the programming language BASIC, and focusses in detail on the development of knowledge about the concept of a program variable, and its use in assignment statements. It suggests how several analogies can be used together to form new concepts where no one analogy would have been sufficient. Errors produced by one reasoning from one analogy can be corrected by another. As an illustration of the main principles of the model, a computer program, CARL, is presented that learns to use variables in BASIC assignment statements. While learning about variables, CARL generates many of the same erroneous hypotheses seen in the recorded protocols of students learning the same material given the same set of analogies. The learning process results in a single target model that retains some aspects of each of the analogies presented. For more information, contact Bernard Silver (617) 576-6212 ------------------------------ Date: 11am 10/22/85 From: Alker@mc Subject: Seminar - Computational Discourse Analysis Using DEREDEC (MIT) [Forwarded from the MIT bboard by SASW@MIT-MC.] Computational Discourse Analysis Using DEREDEC: An Analysis of Balzac's Sarrasine Jaqueline Leon and Jean-Marie Marandin Centre National de la Recherche Scientifique Paris, France We present research in computational discourse analysis and discuss an example for the case of Balzac's Sarrasine. We use P. Plante's DEREDEC programming system in this work because of its suitability for natural language processing. After a bottom-up syntactic parser for French grammar produces a syntactic derivation, we perform pattern matching on the output to achieve a linguistic and literary interpretation. We describe how we use these programs to capture two different aspects of a text: the thematic segmentation and density. Time: 11-12:30, Tuesday, October 22, 1985 Place: Millikan Room, E53-482 Host: Professor Hayward R. Alker, Jr., Department of Political Science, MIT ------------------------------ Date: 18 Oct 85 10:14:51 EDT From: Jeanne.Bennardo@CMU-RI-ISL1 Subject: Seminar - RESEARCHER and Patent Analogies (CMU) Topic: Presentation of RESEARCHER project. Speaker: John C. Akbari Place: DH3313 Date: Wednesday, Oct. 23 Time: 10:00am - 11:00am Speaker: John C. Akbari is a Masters student at Columbia University's Department of Computer Science. He is interested in joining the Intelligent Systems Laboratory's Phoenix project. Below is a description of his artificial intelligence research. Both projects described below investigate different aspects of RESEARCHER, a prototype intelligent information system being developed at Columbia University under the direction of Professor Michael Lebowitz. The domain of investigation is disc drive patents. The result of this research is being implemented in LISP as a component of RESEARCHER. MS Thesis Research involves generating "catalogue descriptions" of hierarchical objects, determining salience as a function of similarity between an instance of an object and the prototype of the object. This will be used in generating information to be passed to a case grammar generator to produce the actual text. We hope to develop a method of determining importance of static information (via "filtering through" the prototype) relative to context. We are studying the interaction of structural, attributive, and functional information on the quality of the description. Further work will investigate the need for different prototypes for different users as an aspect of user modelling, so that a patent lawyer would receive a different description from an engineer, given the same instance. Thesis advisor: Prof. Michael Lebowitz Natural language We are enhancing RESEARCHER's parser to utilize syntactic aspects of relations that cause focus of attention to shift within sentences. This involves modifying memory-based parsing to determine when syntax cues are sufficiently strong to over-ride the need to search memory. Supervisor: Prof. Michael Lebowitz ------------------------------ End of AIList Digest ********************