IRList Digest Tuesday, 29 November 1988 Volume 4 : Issue 57 Today's Topics: Email - USENET version of IRList Interests - Designing lexical databases Query - Articles on comparitive evaluation of IR methods and systems - Hypertext for MSDOS - Connection Machine relevance search and Metamorph Reply - Connection Machine algorithm - Public domain hypertext Call for Papers - REMINDER!! SIGIR 89 papers due soon! CSLI - Interpreting text, Information manipulation, New Ph.D. program COGSCI - Terminological logics, Plausible reasoning News addresses are Internet: fox@vtopus.cs.vt.edu BITNET: foxea@vtcc1.bitnet (replaces foxea@vtvax3) ---------------------------------------------------------------------- From: camco!hpubvwa!vantage!ray@teltone.com (Ray Liere) Newsgroups: comp.theory.info-retrieval Subject: IRList Digest -- Want Them? Message-Id: <1990001@vantage.UUCP> Date: 22 Nov 88 04:42:37 GMT Organization: Vantage Consulting and Research Corporation Lines: 48 Hello readers of the USENET notesgroup "comp.theory.info-retrieval" (and perhaps others who receive this notesgroup via various gateways, interconnects, and other nefarious means). I have been in electronic communication with Ed Fox, the editor of the IRList Digest. There are difficulties in getting the IRList Digests posted so that everyone on USENET gets them. And mailing them to interested persons individually is becoming a monumental and frustrating task for Ed. So ... If you wish, I will post the IRList Digests to this USENET notesgroup. (Ed will first email them to me -- the email connection between Ed's site and ours seems to be reliable ... at least for the time being). Before starting such a process, I am posting this general query to the USENET to see how widespread the non-appearance of the IRList Digests is. In other words, I may be in an isolated pocket and perhaps most of USENET is already getting the IRList Digests ... and would therefore receive two copies of each issue if I were to start posting them. Please post or email or USMail to me if you do NOT want the IRList Digests posted by me to this notesgroup due to your already receiving them. If I do not receive any/many "no" votes, I will post IRList Digests volume 4 issues 43 through 54 (and any others that I receive from Ed in the meantime) on Saturday, 3 December. That should give everyone plenty of time to post or to email/USMail to me. You do not need to email yes votes. (Our uucp connection is over long distance phone lines ...). Besides, I am not doing a formal yes/no tally. Rather, I am trying to avoid making lots of enemies by inadvertently providing IRList Digest to a large number of people in stereo. IMPORTANT: Ed Fox continues to be the editor and moderator of IRList Digest, so ANY communications regarding the IRList Digest should continue to be directed to him. I am only offering to serve as a relay station ... Ray Liere mail: Vantage Consulting and Research Corporation 1017 Molalla Avenue, Suite 3 Oregon City, Oregon 97045, USA voice: (503)657-7294 uucp: hplabs!hpfcla!hpubvwa!vantage!ray [Note: I think this went out over USENET but am including here just in case and also so that current recipients of IRList getting copies directly from me can decide if they wish to work with USENET instead. - Ed.] ------------------------------ Date: Wed, 16 Nov 88 17:00 N From: Subject: membership IRlist Digest, lexical database design I would like to be added to the list of members of IRlist Digest. I am working on the design of a lexical database, so I am interested in information about data bases and methods to set up a large computerized lexicon. With regards, Margret van Gaalen ------------------------------ Date: Tue, 29 Nov 88 09:12:28 EST From: Peter.Capell@CAT.CMU.EDU Subject: Evaluation articles ... What are the seminal articles about comparative evaluation of information retrieval methods and systems? ------------------------------ Date: Fri, 18 Nov 88 11:37 N From: Subject: hypertext In the last issue of Byte there was an article about hypertext. As far as I could find out, the only possibility to get a hypertext system on an IBM (msdos) machine is by using the program GUIDE. This program is hard to get in the Netherlands (version 2.0). Now my questions are: - Is someone working with guide in a hypertext system. - How well does it work (eg with interactive video). - Can somebody tell me where I can buy it. - Are there other possibility's to use hypertext on IBM (msdos). Rob Manders Voorlk2@hwalhw50 [Note: ACM is selling "Hypertext on Hypertext" for IBM-like systems using HyperTies, for Macintosh under HyperCard, and for SUN and Apollo using KMS. There are a lot of vendors in the field and several other choices besides Guide. Guide uses Windows to handle the graphics displays. IBM has Infowindow for working with interactive videodiscs. Window Book, sold by BOX Company, for example, has been around for a while and does not require Windows under MSDOS - we use their software on Virginia Disc One so that even PC/XT systems can be supported. Guide is sold by OWL which is based in UK but has US offices - try 206/747-3203 - Ed.] ------------------------------ Date: Mon, 21 Nov 88 19:11:51 EST From: berlin%bu-albert.BU.EDU@bu-it.bu.edu (David Fickes) Subject: Thinking Machines: Relevance Search Algorithm. I've heard pf the TMC (Connection Machine) Relevance Search Algorithm and all the wonderous things it is supposed to do. Since we now have a connection machine here at BU, I thought I'd look into it. From a few of my contacts, the impression seems to be that it is TMC's implementation of work done by Ingram (?). Essentially, one contact (Thunderstone, EPI - makers of Metamorph) claim it is simply TMC throwing a huge amount of cycles at a problem with mediocre results. Perhaps, one of your folks in the field might be able to render a less obviously biased opinion on the matter? thanks, - david David K. Fickes dfickes@bu-albert.bu.edu The Collected Papers of Albert Einstein ...harvard!bu-it!bu-albert!dfickes Boston University berlin@buita.bu.edu 745 Commonwealth Avenue - room 541 617/ 353-9249 Boston, MA 02215 617/ 783-4301 [Note: I have seen the TMC system and they has a presentation at ASIS 88 recently. Salton in CACM recently wrote about their work with SUN systems that gave in many cases better results and did not require large hardware investment - but of course there are fewer potential simultaneous users. Regarding Metamorph, I have still not heard any scientific descriptions about what they do that would enable me to comment on their system. - Ed] ------------------------------ Date: 23 Nov 88 11:29 PST From: Pedersen.pa@Xerox.COM Subject: Re: Thinking Machines: Relevance Search Algorithm. Essentially, The Connection Machine algorithm implements a keyword based relevance feedback search without inverted indices. The entire corpus is represented in core as a series of hash tables, one for each document. Each document is queried in parallel to determine if a particular word from the reference document is contained within it. If a hit occurs, that document's score is incremented by some weight (inverse frequency weight, I believe). After each keyword in the reference document has been processed, the documents are sorted according to their scores --- and some number of the highest rated documents are returned as the mostly likely "relevant" candidates. Thinking Machines reports high preformance for this strategy, although a fallback scheme is required for cases in which the corpus is too large to fit into core (more than 256 Megabytes compressed). I recall seeing an article written by Salton stating that roughly similar preformance can be achieved by precomputing an inverted index. J.P. ------------------------------ Date: Wed, 16 Nov 88 12:28:12 DNT From: Jakob Nielsen Tech Univ of Denmark Subject: Re: Looking for public domain hypertext I have a shareware hypertext which you could get. It is developed under HyperCard and includes features such as a history list, dynamic ("pseudo") overview maps, user backtracking, footprinting, time-since-you-were-here notification, etc. The subject matter of this hypertext is a report from the HyperTEXT'87 workshop. Note that the current version of this document is vers. 3.0 (vers. 1 and 2 used other interaction techniques). You can get it by sending US dollars 15 in a check made out to Jakob Nielsen to the following address: Jakob Nielsen Department of Computer Science Technical University of Denmark Building 344 DK-2800 Lyngby Copenhagen DENMARK Tel: International access +45-1-38 23 20 Email: datJN@NEUVM1.bitnet Fax: International access +45-2-88 22 39 You are also welcome to copy this document for free if you can find somebody who has a copy, since I am not actually *selling* it but just charging a fee for sending out copies (since I have had far too many requests to handle for free). ------------------------------ Date: Tue, 29 Nov 88 11:03:24 EST From: Edward A. Fox Subject: SIGIR 89 Call for Papers REMINDER! In V4 #36 I had the Call for Papers for SIGIR 89. The deadlines are fast approaching for submissions: 14 Dec. for papers and panel proposals 16 Dec. for tutorial proposals and we have not received many submissions yet. Please remember to send in your papers for this exciting conference! - Ed Fox (Publicity Chair) ------------------------------ Date: Wed, 16 Nov 88 16:46:11 PST From: Emma Pease Subject: CSLI Calendar, November 17, 4:9 [Extract - Ed.] ... NEXT CSLI SEMINAR The Resolution Problem for Natural-Language Processing Week 9: Interpretation as Abduction Jerry Hobbs (hobbs@ai.sri.com) December 1 We will return to a discussion of knowledge-based AI approaches to the resolution problem, and in particular to an approach using a scheme for abductive inference developed in the TACITUS project at SRI. It will be argued that to interpret a text, one must prove the logical form of the text from what is already mutually known, merging redundancies where possible and making assumptions where necessary. It will be shown how the problems of, among others, reference, ambiguity, and metonymy can be addressed with this method. This approach, in addition, suggests an elegant and thorough integration of syntax, semantics, and pragmatics---one moreover that works for integration and generation both. This will be described, and its significance for modularity will be discussed. ____________ STASS SEMINAR Multimodal, Information-based Inference Jon Barwise, Alan Bush, and John Etchemendy (barwise@csli.stanford.edu, bush@csli.stanford.edu, etch@csli.stanford.edu) Cordura Conference Room December 1, 4:00-5:30 We will talk about our work designing an inference system that allows the direct manipulation of information provided via different modalities (e.g., visual and sentential). We will demonstrate a mock-up of a program we are developing to teach this approach to inference. Time and place subject to change due to the availablity of equipment. ____________ ANNOUNCEMENT The Stanford Department of Philosophy announces a new special program within their Ph.D. program: Philosophy and Symbolic Systems. The program is designed to allow students to do interdisciplinary coursework and research in the area of symbolic systems. For more information, contact the philosophy department (723-2547) or Jon Barwise (barwise@csli.stanford.edu). ------------------------------ Date: Mon, 28 Nov 88 13:51:00 EST From: Peter de Jong Subject: Cognitive Science Calendar [Extract - Ed.] Reply-To: cog-sci-request@WHEATIES.AI.MIT.EDU Return-Path: Date: Mon 21 Nov 88 17:40:09-EST From: Marc Vilain Subject: BBN AI Seminar: Peter F. Patel-Schneider BBN Science Development Program AI Seminar Series Lecture COMPLEXITY AND DECIDABILITY OF TERMINOLOGICAL LOGICS Peter F. Patel-Schneider AI Principles Research Department AT&T Bell Laboratories (pfps@allegra.att.com) BBN Labs 10 Moulton Street 2nd floor large conference room 10:30 am, Tuesday November 29 Terminological Logics are important formalisms for representing knowledge about concepts and objects, and are attractive for use in Knowledge Representation systems. However, Terminological Logics with reasonable expressive power have poor computational properties, a fact which has restricted their use and utility in Knowledge Representation systems. This talk gives a brief description of Terminological Logics, presents some results concerning their tractability and decidability, and discusses the role of Terminological Logics in Knowledge Representation systems. ------------------------------ Date: Tue, 22 Nov 88 15:08:38 EST From: sarah@wheaties.ai.mit.edu (Sarah A. Shellard) Subject: seminar on November 29, 3:00pm New time : 3:00pm Date : Tuesday, November 29 1988 Topic : THEORY OF PLAUSIBLE REASONING and TWO-TIERED KNOWLEDGE REPRESENTATIONS Speaker : Professor Ryszard S.~Michalski Center for Artificial Intelligence George Mason University Abstract : A new cognitive theory of plausible reasoning (developed by A.~Collins and the author) will be described and illustrated by examples. The recently introduced idea of two-tiered representation of imprecise and context-dependent concepts will also be discussed, and other major projects conducted at the GMU Center for Artificial Intelligence will be outlined. Most of the current theories of uncertain reasoning are prescriptive, and concerned primarily with combining the certainties of the premises to determine the certainty of the conclusion. In contrast, the presented theory is descriptive, and concentrates on structural properties of knowledge and on rules of inference that support plausible reasoning. The theory tries to explain how people actually reason in the presence of uncertainty. Various plausible reasoning rules will be presented that are based on specialization, generalization, similarity and dissimilarity operations. The theory will be illustrated by a result from applying it to a problem of predicting properties of an unknown chemical element. The idea of two-tiered concept representation deals with the problem of how to represent and learn imprecise and context-dependent concepts (such concepts are called flexible). It postulates that flexible concepts should be represented by two components: The Base Concept Representation (BCR) and the Inferential Concept Interpretation (ICI). The BCR captures the essential and typical concept properties, and the ICI uses various inference rules and meta-rules for handling the concept variability, exeptional cases and context-dependency. Early experimental results have demonstrated that such a representation can lead to a substantial reduction of memory requirements for concept representation and, at the same time, to an improvement of the performance in recognizing new instances. Refreshments : 2:45 pm ------------------------------ END OF IRList Digest ********************