IRList Digest Thursday, 10 December 1987 Volume 3 : Issue 46 Today's Topics: Query - Detecting language from a title - Hypertext bibliography? - References to knowledge based tools for AI bibliographies Interest - History and retrieval - IR analyst using SPIRES Announcement - Impact of new technology on information professionals COGSCI - Grouping in recognition - Comparitive analysis, Truth and cognitive science, Unified Medical Language system News addresses are Internet or CSNET: fox@vtopus.cs.vt.edu BITNET: foxea@vtvax3.bitnet ---------------------------------------------------------------------- Date: 30-NOV-1987 19:17:17 GMT From: LOU@VAX.OXFORD.AC.UK Subject: an interesting problem Here's an interesting problem someone may have an answer to: what's the best way of automatically detecting the language in which something is written? We have a library here in Oxford with a large (well, very large actually) catalogue of book titles in just about every european language you can think of: english greek latin german hebrew french russian... in order to get the indexing strategy right (it's a bit dim to mark "the" as a stop word if the title is in French) to say nothing of the hyphenation points, it would be nice to get each title tagged by its language. As there are something like one and a quarter million titles (I did say it was large) it would be even nicer to do this at least semi-automtically. Any suggestions? High frequency words might be one possibility, except that titles are mostly (but not all) quite short. Has anyone done anything similar with trigrams? Lou Burnard (LOU @ UK.AC.OXFORD.VAX ) ------------------------------ Date: Tue, 1 Dec 1987 11:00 EST From: James Nolte <$JSN@CLVM> Subject: Brown Bibliography on Hypertext from Hypertext 87 You mentioned in the IRLIST Digest of Wednesday, 25 November that Brown distributed a bibliography on Hypertext. Is that Brown University? Do you have an address or contact person from whom I could obtain such a bibliography? [Note: try ny@iris.bitnet which will get you to the IRIS Project at Brown. Maybe we will get some news from IRIS soon? - Ed] ------------------------------ Date: Wed, 2 Dec 87 08:59 EST From: Roland Zito-Wolf Subject: Knowledge-based bibliographies I am looking for references regarding knowledge-bases and KB-based tools for organizing a bibliographic database on AI. I want to be able to retrieve references by various indices. Specific issues I'd like to know about: - friendly data entry - searching through alternate paths (say, finding articles related to a given article in some way: by author, topic, system name, etc.) - ability to "evolve" the structure of the KB with time - what is areasonable conceptual structure for reference databases, in general? I'll post a digest of responses to the list. Roland J. Zito-wolf Palladian Software 4 Cambridge Center Cambridge, Mass 02142 617-661-7171 RJZ%JASPER@LIVE-OAK.LCS.MIT.EDU ------------------------------ Date: Thu, 3 Dec 87 09:40:12 GMT From: F.E.Candlin@VME.GLASGOW.AC.UK Subject: SUBSCRIPTION TO IRList Dear Professor Fox, This note is to ask you if you would be willing to put me down for subsription to IRList. I work as the programmer at the DISH History and Computing Lab at Glasgow University, Scotland. At Glasgow, we have a number of fairly large databases - bankruptcies, marriage records, property valuations, trade accounts etc - which were produced largely using software developed by ourselves. In addition about half of the academic staff involved in using computers in history teaching are now producing smaller datasets specifically related to their courses. One of the interesting things about this kind of data is the fact that it usually existed in retrievable form long before computers were invented. To maintain the integrity of the source, we usually try to reflect the original form of the data in our databases (be it the marriage register, the account book or whatever). Very few dbms's cope well with such data - normally preferring explicit relationships between items of interest and disliking small inconsistencies whem carrying out analysis. More strain is put on the dbms by the demands of the typical historian, who is as interested in exceptions as general statistics. He is also interested in following up specific people who may pop up in a number of otherwise unrelated sources. None of this is unique to historians, of course, but the insight that we have gained has inspired us to set up a new project to look into ways of instructing information managers in the implications of large- sets of messy data. My address is: F.E.Candlin, DISH History Computing Laboratory, 2 University Gardens, Glasgow University, Glasgow G12 8QQ, SCOTLAND Telephone: 041 339 8855 x 4510 Email: F.E.Candlin@UK.AC.Glasgow.VME Many Thanks ------------------------------ Date: Fri, 27 Nov 87 11:50:02 MST From: Terry Butler Subject: Join IRList I would like to join IRList. I am an IR analyst in the computing department at the University of Alberta, Edmonton, Alberta, Canada. Our unit provides computing support for academics at our university. Our major offering is SPIRES on the mainframe; and we are supporting several micro data base packages. ------------------------------ Date: Sun, 29 Nov 87 20:05:19 EST From: dws@EDDIE.MIT.EDU (Don W. Saklad) Message-Id: <8711300105.AA11893@EDDIE.MIT.EDU> To: FOX@VTCS1.BITNET Subject: notice posted at our local public library Summary: interesting notice posted on our local public library bulletin board Keywords: library libraries The Graduate School of Library & Information Science Simmons College invites you to attend a lecture in The Samuel Lazerow Memorial Lecture Series sponsored by The Institute for Scientific Information featuring Dr. Elizabeth Young Vice-President, INMARSAT, Policy and Representation COMSAT Maritime Services An executive in one of the nations's leading satellite communications companies, Dr. Young will share her insights: "the impact of new technologies on the role and status of the information professional." Wednesday, December 9, 1987 Simmons College Auditorium 300 The Fenway Boston Schedule of Events ---------------------- 2:00-2:30 Registration 2:30-4:00 Welcome and Introduction Dean Robert D. Stueart Lazerow Lecture Dr. Elizabeth Young 4:00-5:00 Reception and refreshments, GSLIS Lounge Open House, GSLIS Computer Access Laboratory Hardware and software recently acquired by the School will be on display in the Computer Access Laboratory. Faculty and students will be available to discuss and demonstrate some of the new technologies which include BiblioFile, SilverPlater, Dissertation Abstracts on Disc, DATEXT, Dialog on Disc, OCLC's Search CD-450, WilsonDisc and the Library Students Association electronic bulletin board. Attendance by reservation only. RSVP to Linda Willey, at 617-738-2223 by Monday December 7 [Note: Can anyone tell us what happened? - Ed] ------------------------------ Date: Mon, 16 Nov 1987 13:17 EST From: Peter de Jong Subject: Cognitive Science Calendar [Extract - Ed] Date: Saturday, 14 November 1987 11:36-EST From: Paul Resnick Re: AI Revolving Seminar Thursday 19, November 4:00pm Room: NE43- 8th floor Playroom The Artificial Intelligence Lab Revolving Seminar Series THE USE OF GROUPING IN VISUAL OBJECT RECOGNITION David Jacobs Many systems have been developed for recognizing two and three dimensional objects in images. Some problems emerge, however, when we try to extend these approaches to handle more complex tasks. More complex tasks might include using knowledge of large libraries of different objects instead of looking for just a single object, knowing about flexible objects instead of rigid ones, or recognizing objects in large, complex images. All these tasks require much more computation from existing recognition systems, and make them more prone to commit errors. This talk will describe an object recognition system which attempts to deal with these problems of accuracy and complexity by using grouping. The system, called GROPER, knows about a library of different two-dimensional objects, and examines images which contain some of these objects, perhaps partially occluded. It proceeds by forming groups of image edges which seem particularly likely to have come from a single object. It then matches these groups of image edges to groups of model edges by hashing. Grouping allows GROPER to first try to perform recognition with the groups of edges most likely to lead to the correct recognition of an object. A comparison between GROPER's performance and that of a similiar recognition system which does not use grouping shows that this can dramatically reduce the amount of computation required for recognition and dramatically reduce the number of mistakes made. ------------------------------ Date: Sat, 28 Nov 1987 12:40 EST From: Peter de Jong Subject: Cognitive Science Calendar [Extract - Ed] Date: Tuesday, 24 November 1987 18:22-EST From: Marc Vilain Re: BBN AI Seminar -- Daniel Weld BBN Science Development Program AI Seminar Series Lecture THEORIES OF COMPARATIVE ANALYSIS Daniel S. Weld MIT Artificial Intelligence Lab (WELD@REAGAN.AI.MIT.EDU) BBN Labs 10 Moulton Street 2nd floor large conference room 10:30 am, Tuesday December 1 This talk analyzes two approaches to a central subproblem of automated design, diagnosis, and intelligent tutoring systems: comparative analysis. Comparative analysis may be considered an analog of qualitative simulation. Where qualitative simulation takes a structural model of a system and qualitatively describes its behavior over time, comparative analysis is the problem of predicting how that behavior will change if the underlying structure is perturbed and also explaining why it will change. For example, given Hooke's law as the model of a horizontal, frictionless spring/block system, qualitative simulation might generate a description of oscillation. Comparative analysis, on the other hand, is the task of answering questions like: ``What would happen to the period of oscillation if you increase the mass of the block?'' I have implemented, tested, and proven theoretical results about two different techniques for solving comparative analysis problems, differential qualitative (DQ) analysis and exaggeration. DQ analysis would answer the question above as follows: ``Since force is inversely proportional to position, the force on the block will remain the same when the mass is increased. But if the block is heavier, then it won't accelerate as fast. And if it doesn't accelerate as fast, then it will always be going slower and so will take longer to complete a full period (assuming it travels the same distance).'' Exaggeration can also solve this problem, but it generates a completely different answer: ``If the mass were infinite, then the block would hardly move at all. So the period would be infinite. Thus if the mass was increased a bit, the period would increase as well.'' Both of these techniques has advantages and limitations. DQ analysis is proven sound, but is incomplete. It can't answer every comparative analysis problem, but all of its answers are correct. Because exaggeration assumes monotonicity, it is unsound; some answers could be incorrect. Furthermore, exaggeration's use of nonstandard analysis makes it technically involved. However, exaggeration can solve several problems that are too complex for DQ analysis. The trick behind its power appears to have application to all of qualitative reasoning. ============================== Date: Thursday, 19 November 1987 12:09-EST From: Eric Sven Ristad Tuesday, 1 December 7:30pm Room: 34-401 (Grier Conference Room) TRUTH AND COGNITIVE SCIENCE Hilary Putnam Department of Philosophy, Harvard University The following facts are commonly cited as examples of "intentionality": (i) the fact that words, sentences, and other "representations" have meaning; (ii) the fact that representations may refer to some actually existing thing or each of a number of actually existing things; (iii) the fact that representations may be about something which does not exist; and (iv) the fact that a state of mind may have a "state of affairs" as its object, as when someone says, "she believes that [he is trustworthy]." When the computer revolution burst upon the world, it was widely expected (and claimed) that computer models would explain the nature of these various phenomena. In short, people expected that a reductive account of the various topics included under the chapter-heading "intentionality" would be given. Now that this has not proved so easy, a number of thinkers are beginning to suggest that it isn't so bad if this can't be done; intentionality is only a feature of "folk psychology" anyway. If a first class scientific account of intentional facts and phenomena can't be given, that is not because scientific reductionism is not the right line to take in metaphysics, but rather it is because there is, so to speak, nothing there to reduce. I want to argue that both attitudes are mistaken; that intentionality won't be reduced and won't go away. Commentary: Jerry Fodor Department of Philosophy, CUNY Copies of paper Karen persinger, 20B-225, 253-7358 ============================== Date: Tuesday, 24 November 1987 12:34-EST From: Rosemary B. Hegg Re: Komorowski Seminar DATE: Wednesday, December 2, 1987 TIME: Refreshments: 1.45PM Lecture: 2.00PM PLACE: NE43-8th floor playroom THE UNIFIED MEDICAL LANGUAGE SYSTEM HENRYK JAN KOMOROWSKI Decision Systems Laboratory Harvard Medical School/Brigham and Women's Hospital MIT Artificial Intelligence Laboratory Knowledge systems in medical applications had several undeniable successes, yet it is fair to say that only very few of the systems found their way to everyday use. One well recognized impediment to progress is the lack of a knowledge-base (KB) which would encompass a broad spectrum of medical knowledge. Instead of embarking on a 200 man/year project to encode a comprehensive KB of modern medicine, a group of academic institutions joined efforts and expertise to develop a canonical taxonomy of medical terms and relations. This canonical taxonomy, called a Unified Medical Language System, will provide standards for structuring, indexing, retrieving, and communicating medical knowledge. It is anticipated that the UMLS will be the kernel of most future biomedical applications. An application may develop its knowledge-base as an extension to the taxonomy and use the UMLS as an interlingua to communicate with other applications. In this scenario parallel development and sharing of multiple expert resources will be possible. One critical issue in the creation of UMLS and a particular focus in our research is the identification of semantic features and relations which should be represented in the taxonomy, and the design of appropriate structures and tools for storing, displaying, and authoring these features and relations. The currently developed prototypical taxonomy and the viewing and authoring environment has begun to open the road to a magnitude of applications. They include an environment for learning the structure of medicine, efficient preparation of queries to the external body of the existing medical literature, automatic acquisition of medical knowledge, automatic identification of related concepts, free browsing in pursuit of curiosity, etc. The development of the UMLS continues to challenge both the computer scientists and the medical community. HOST: Peter Szolovits ------------------------------ END OF IRList Digest ********************