LAWS@SRI-AI.ARPA (02/11/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI.ARPA> AIList Digest Monday, 11 Feb 1985 Volume 3 : Issue 19 Today's Topics: Machine Translation - Slocum's System, Publications - Manual of Intensional Logic, Seminars - Algebraic Specifications (CSLI) & Massively Parallel Natural Language Processing (BBN) & Belief, Awareness, and Limited Reasoning (CSLI) & The Logical Data Model (SU), Conferences - Evolution and Information & SCCGL Conference on General Linguistics ---------------------------------------------------------------------- Date: Sun 10 Feb 85 17:35:14-PST From: LOUROBINSON@SRI-AI.ARPA Subject: Machine Translation Regarding a German/English translator: How about Jonathan Slocum's program developed for Siemens? Slocum is now at MCC in Austin courtesy of the University of Texas. Good luck. Lou Robinson ------------------------------ Date: Wed 6 Feb 85 17:20:59-PST From: Emma Pease <Emma@SU-CSLI.ARPA> Subject: Lecture Notes - Manual of Intensional Logic [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] FIRST VOLUME OF CSLI LECTURE NOTES The first in the series of CSLI Lecture Notes has just been published. Entitled ``Manual of Intensional Logic,'' the 75-page book by Johan van Benthem constitutes a graduate course that the author taught in the Winter of 1984 while at CSLI. ``Intensional Logic as understood here,'' the author writes in the Introduction, ``is a research program based upon the broad presupposition that so-called `intensional contexts' in natural language can be explained semantically by the idea of `multiple reference.' '' Unlike CSLI Reports, the Lecture Notes will be sold for a nominal fee to defray part of production costs. The price of ``Manual of Intensional Logic'' is $5, and it may be purchased at the Stanford Bookstore or by writing to Dikran Karagueuzian at the Center. A 25% discount is offered to all members of the CSLI community or to anyone ordering three or more copies to be used for instructional purposes. California residents should add sales tax. ------------------------------ Date: Wed 6 Feb 85 17:20:59-PST From: Emma Pease <Emma@SU-CSLI.ARPA> Subject: Seminar Summary - Algebraic Specifications (CSLI) [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] SUMMARY OF AREA C MEETING ``Algebraic Specifications in an Arbitrary Institution'' Andrzej Tarlecki Dept. of Computer Science, Univ. of Edinburgh The pioneering papers on algebraic specification used many-sorted equational logic as a logical framework in which specifications were written and analyzed. Nowadays, however, examples of logical systems in use include first-order logic, higher-order logic, infinitary logic, temporal logic, and many others. Note that all these logical systems may be considered with or without predicates, admitting partial operations or not. This leads to different concepts of signature and of model, perhaps even more obvious in examples like polymorphic signatures, order-sorted signatures, continuous algebras, or error algebras. The informal notion of a logical system for writing specifications has been formalized by Goguen and Burstall who introduced for this purpose the notion of institution. The first and presumably most important application of this notion is its use in the theory of algebraic specifications. It turns out that most of the work on algebraic specification, especially concerning specification languages, may be done in an institution-independent way. We briefly present a collection of simple but very powerful specification-building operations and give their semantics in an arbitrary institution. In this context we outline a very simple and mathematically elegant view of the formal development of programs from their specifications. The notion of institution is also used to formulate (and prove) some model-theoretic results at an appropriately general level. We show how to generalize to an arbitrary institution a Birkhoff-type characterization of quasi-varieties as implicational classes. This result may be used to prove that Mahr and Makowsky's characterization of standard algebraic institutions which strongly admit initial semantics holds for arbitrary institutions satisfying a number of technical assumptions. Finally, we briefly outline some problems concerning the notion of institution itself. We discuss the need for some tool for constructing new institutions and for combining institutions (``putting institutions together''). We also indicate possible generalization of this notion which would provide a mold for richer semantical systems than just collections of sentences with a notion of their truth. ------------------------------ Date: 8 Feb 1985 14:58-EST From: Brad Goodman <BGOODMAN@BBNG> Subject: Seminar - Massively Parallel Natural Language Processing (BBN) [Forwarded from the MIT bboard by JCMA@MIT-MC.] Massively Parallel Natural Language Processing Professor David L. Waltz, Thinking Machines and Brandeis University Date: 10:00 a.m. Tuesday, February 19, 1985 Place: Newman Auditorium BBN Laboratories Inc. 70 Fawcett Street Cambridge, Ma. This talk will describe research in developing a natural language processing system with modular knowledge sources but strongly interactive processing. The system offers insights into a variety of linguistic phenomena and allows easy testing of a variety of hypotheses. Language interpretation takes place on an activation network which is dynamically created from input, recent context, and long-term knowledge. Initially ambiguous and unstable, the network settles on a single interpretation, using a parallel, analog relaxation process. The talk will also describe a parallel model for the representation of context and of the priming of concepts. Examples illustrating contextual influence on meaning interpretation and "semantic garden path" sentence processing, along with a discussion of the building and implementation of a large scale system for new generation parallel computers are included. ------------------------------ Date: Wed 6 Feb 85 17:20:59-PST From: Emma Pease <Emma@SU-CSLI.ARPA> Subject: Seminar - Belief, Awareness, and Limited Reasoning (CSLI) [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] 12 noon, 2/14 TINLunch Ventura Hall ``Belief, Awareness, and Limited Reasoning'' Conference Room Ronald Fagin, IBM San Jose Research Laboratory Possible-worlds semantics for knowledge and belief do not seem appropriate for modelling human reasoning since they suffer from the problem of what Hintikka calls ``logical omniscience''. This means that agents are assumed to be so intelligent that they must, in particular, know all valid formulas. Moreover, each agent's knowledge is also closed under deduction, so that if an agent knows p, and if p logically implies q, then the agent must also know q. Unfortunately, this is certainly not a very accurate account of how people operate! People are not logically omniscient for several reasons, including (1) Lack of awareness: how can someone say that he knows or doesn't know about p if p is a concept he is completely unaware of? (2) People are resource-bounded: they simply lack the computational resources to deduce all the logical consequences of their knowledge. (3) People don't focus on all issues simultaneously: it is possible for a person to have distinct frames of mind, where the conclusions drawn in distinct frames of mind may contradict each other. Some new logics for belief and knowledge are introduced which model these phenomena, so that, in particular, agents need not be logically omniscient. This talk represents joint work with Joe Halpern. --Ronald Fagin ------------------------------ Date: Tue, 5 Feb 85 09:08:22 pst From: Gabriel Kuper <kuper@diablo> Subject: Ph. D. Oral - The Logical Data Model (SU) [Forwarded from the Stanford bboard by Laws@SRI-AI.] Gabriel M. Kuper The Logical Data Model: A New Approach to Database Logic 9AM, 7 Feb. 985 Building 420 (Psychology), Room 41 We propose a mathematical framework for unifying and generalizing the three principal data models, i.e., the relational, hierarchical and network models. Until recently most theoretical work on databases has focused on the relational model, mainly due to its elegance and mathematical simplicity compared to the other models. Some of this work has pointed out various disadvantages of the relational model, among them its lack of semantics and the fact that it forces the data to have a flat structure that the real data does not always have. The Logical Data Model (LDM) combines the advantages of both approaches. It models database schemas as directed graphs, in which the leaves correspond to the attributes, and the internal nodes to connections between the data. Instances of LDM schemas consist of r-values, which constitute the data space, and l-values, which constitute the address space. This enables us to deal with instances of cyclic structures, but still get a first-order theory. We define a logic on LDM schemas in which integrity constraints can be specified, and use it to define a logical, i.e. non-procedural, query language that is analogous to Codd's relational calculus. We also describe an algebraic, i.e. procedural, query language and prove that the two query languages are equivalent. These languages have a novel feature: not only can they access a non-flat data structure, e.g. a hierarchy, but the answers they produce do not have to be flat either. Thus, the language really does have the ability to restructure data and not only to retrieve it. ------------------------------ Date: Wed 6 Feb 85 17:20:59-PST From: Emma Pease <Emma@SU-CSLI.ARPA> Subject: Conference on Evolution and Information [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] CONFERENCE ON EVOLUTION AND INFORMATION A conference on Evolution and Information with major support from CSLI will be held at Stanford this April 19-21. The specific focus of the conference will be on the use of optimality models both in biology and in the human sciences. Papers will be contributed to the conference by biologists, philosophers, psychologists, and anthropologists. Apart from addressing problems and limitations of optimality models within biology, an important aim of the conference will be to explore the relevance of biological results, either factually or methodologically, to other areas of inquiry. Papers to be discussed at the conference will be circulated about a month before the meeting. Contributors will be asked to give a brief summary of their papers at the conference sessions but papers will not be read. Therefore, anyone who would be interested in seeing the papers in advance, or would like any further information about the conference, should contact John Dupre, Philosophy, Stanford University (415-497-2587, Dupre@Turing). ------------------------------ Date: Tue, 5 Feb 85 20:13:24 pst From: li51x%sdcc3@SDCSVAX Subject: SCCGL Conference on General Linguistics CALL FOR PAPERS SOUTHERN CALIFORNIA CONFERENCE ON GENERAL LINGUISTICS APRIL 20-21, 1985 The Conference will be held at the University of California, San Diego. Papers from any of the subdisciplines of linguistics are eligible. Graduate students are especially encouraged to participate, and abstracts will be refereed anonymously. Please provide 10 copies of your single-page titled anonymous abstract, and include an index card with the fol- lowing information: Paper title (matching that on abstract) Author Address Phone number (including area code) Please send abstracts to the address below before 28 February 1985. Chilin Shih SCCGL Linguistics, C-008 UCSD La Jolla, CA 92093 Information about meals and accommodation will be mailed later. For further information call (619) 452-3600, Chilin Shih, Carol Georgopoulos, or Diane Lillo-Martin. You may reach Chilin at sdcc6!ix226@UCSD.arpa. Please use SCCGL as the subject heading. ------------------------------ End of AIList Digest ********************