nl-kr-request@cs.rpi.edu (NL-KR Moderator Chris Welty) (05/05/89)
NL-KR Digest (Thu May 4 16:26:44 1989) Volume 6 No. 25 Today's Topics: CALL FOR IJCAI-89 STUDENT VOLUNTEERS IJCAI-89 Blackboard Workshop (Revised dates, please post) CSLI Calendar, 4 May, 4:25 Roger Ratcliff AI Seminar Submissions: nl-kr@cs.rpi.edu Requests, policy: nl-kr-request@cs.rpi.edu Back issues are available from host archive.cs.rpi.edu [128.213.1.10] in the files nl-kr/Vxx/Nyy (ie nl-kr/V01/N01 for V1#1), mail requests will not be promptly satisfied. If you can't reach `cs.rpi.edu' you may want to use `turing.cs.rpi.edu' instead. ----------------------------------------------------------------- To: nl-kr@cs.rpi.edu Date: Thu, 4 May 89 10:46:58 EDT >From: Robert_Lindsay@ub.cc.umich.edu Subject: CALL FOR IJCAI-89 STUDENT VOLUNTEERS CALL FOR IJCAI-89 STUDENT VOLUNTEERS The 1989 International Joint Conference on Artificial Intelligence will be held in Detroit, Michigan USA from Saturday, August 19 to Friday August 25. Student volunteers are being sought to assist with a variety of tasks at the conference site, such as registration, ticket taking, messages, and answering questions. Each volunteer will be expected to work 12 hours, and will be provided with free conference registration and Proceedings. If you are interested in volunteering you will need to supply the following information: (1) Name and address (2) An e-mail address if available (3) Name of your university and department, and your level of study (4) The days you prefer to work, jobs desired, or sessions you prefer to assist with (5) Identify any language other than English in which you can converse. Send this information to Robert_Lindsay@UB.CC.UMICH.EDU (Note the underline character between names) E-mail will be acknowledged and further details supplied. If you do not receive a reply within one week, or if you do not have access to e-mail, send the information by regular mail to Prof. Robert Lindsay MHRI University of Michigan 205 Washtenaw Place Ann Arbor, Michigan 48109 ------------------------------ To: nl-kr@cs.rpi.edu Date: Mon, 1 May 89 13:00 EST >From: "Dan Corkill, COINS, UMass 413/545-0156" <CORK@cs.umass.EDU> Subject: IJCAI-89 Blackboard Workshop (Revised dates, please post) - ----------------------------------------------------------------------------- CALL FOR PARTICIPATION IJCAI-89 Workshop on Blackboard Systems Sponsored by AAAI Wednesday, August 23, 1989 Detroit, Michigan, U.S.A. **** PLEASE NOTE NEW SUBMISSION DEADLINE **** Description ----------- The blackboard paradigm is a powerful technique for implementing today's ambitious AI applications and for integrating diverse problem solving expertise into a common framework. The Third Annual Workshop on Blackboard Systems, like its predecessors, provides an informal forum where researchers in blackboard technology and developers of blackboard-based applications exchange ideas, experiences, problems, and inspirations. The aims of the workshop include: allowing participants to share the latest results of their research in an informal setting, informing participants of other researchers working on similar problems or using similar approaches, and identifying common unsolved research issues. Topics ------ Topics of interest for the workshop include: * blackboard systems/shells; * blackboard control mechanisms/techniques; * real-time, parallel, and distributed blackboard approaches; * performance measures for blackboard systems/applications; * user interfaces/explanation facilities for blackboard systems; * application development/debugging facilities for blackboard systems; * problems associated with fielding a blackboard-based application; * novel blackboard-based applications. Submissions presenting comparison data between blackboard technology and other AI methodologies or among alternate blackboard-based approaches are particularly encouraged. Format ------ The workshop is one-day long and will take place on Wednesday, August 23. Accepted papers will be grouped into three panels based on content. Each panel will consist of a series of informal paper presentations followed by a general discussion period. A chair for each panel will be selected from members of the Workshop Committee. As with previous years, a proceedings containing complete versions of the accepted papers will be distributed at the workshop. Submission Information ---------------------- Workshop invitations will be issued on the basis of extended abstracts 10 pages or less in length. Each extended abstract will be reviewed by members of the Workshop Committee. At most, 2 invitations will be issued for each accepted abstract. In keeping with an informal workshop, the total number of invitations will be limited to 30--35 people. Workshop Committee ------------------ Larry Baum, Boeing ATC Kevin Gallagher, UMass (co-chair) Roberto Bisiani, CMU Barbara Hayes-Roth, Stanford Daniel Corkill, UMass (chair) V. Jagannathan, American Cimflex Raj Dodhiawala, FMC Victor Lesser, UMass Robert Engelmore, Stanford Penny Nii, Stanford Lee Erman,Teknowledge Important Dates --------------- May 31, 1989 Extended abstracts must be received July 17, 1989 Notification of invitation or rejection August 8, 1989 Completed papers must be received August 23, 1989 Workshop date Send four copies of extended abstracts to: ------------------------------------------ Daniel D. Corkill Department of Computer and Information Science University of Massachusetts Amherst, Massachusetts 01003 U.S.A. Internet: Cork@CS.UMass.Edu Phone: 413/545-0156 FAX: 413/545-1249 - ----------------------------------------------------------------------------- ------------------------------ To: nl-kr@cs.rpi.edu Date: Wed, 3 May 89 17:16:46 PDT >From: emma@csli.Stanford.EDU (Emma Pease) Subject: CSLI Calendar, 4 May, 4:25 C S L I C A L E N D A R O F P U B L I C E V E N T S _____________________________________________________________________________ 4 May 1989 Stanford Vol. 4, No. 25 _____________________________________________________________________________ A weekly publication of The Center for the Study of Language and Information, Ventura Hall, Stanford University, Stanford, CA 94305 ____________ CSLI ACTIVITIES FOR THIS THURSDAY, 4 May 1989 12:00 p.m. TINLunch Cordura Hall Machine Translation Conference Room Annie Zaenen (zaenen.pa@xerox.com) Abstract in last week's Calendar 2:15 p.m. CSLI Seminar Cordura Hall Varieties of Context: Session 5 Conference Room Language Use in Context: How Does it Work? (or What's Context Good For, Anyway?) Susan Stucky (stucky.pa@xerox.com) Respondent: Herb Clark Abstract in last week's Calendar 3:30 p.m. Tea Ventura Hall 4:00 p.m. STASS Seminar Cordura Hall Beyond Formality: Progress Report on the Conference Room Foundations of Computation Brian Smith (briansmith.pa@xerox.com) No abstract ____________ CSLI ACTIVITIES FOR NEXT THURSDAY, 11 May 1989 12:00 p.m. TINLunch Cordura Hall Information Theory, Statistical Inference, Conference Room and the Boot Strap Brad Efron, Department of Statistics, Stanford Abstract in next week's Calendar 2:15 p.m. CSLI Seminar Cordura Hall Varieties of Context: Session 6 Conference Room Indexicality in the Substrate Brian Smith (briansmith.pa@xerox.com) Respondent: John Perry Abstract below 3:30 p.m. Tea Ventura Hall 4:00 p.m. STASS Seminar Cordura Hall Kris Halvorsen Conference Room (halvorse@arisia.xerox.com) Abstract in next week's Calendar ____________ NEXT WEEK'S CSLI SEMINAR Varieties of Context: Session 6 Indexicality in the Substrate Brian Smith (briansmith.pa@xerox.com) Respondent: John Perry May 10 It is sometimes thought that indexicality and context-dependence are complicating features of language and mind---that constants or fixed names are simpler, and in that sense a more elementary case. In this talk I will argue for almost exactly the opposite conclusion. In particular, I will attempt to demonstrate that a certain kind of proto-indexicality "comes for free" from the physical substrate, and should therefore be treated as the base case. From this point of view, it looks as if the real semantic phenomenon that needs explaining is the rise of context-"independence". ____________ LINGUISTICS DEPARTMENT COLLOQUIUM Discrepancies between Comprehension and Production Implications for Acquiring and Representing Linguistic Knowledge Eve Clark (eclark@psych.stanford.edu) Friday, 5 May, 3:30 Cordura Conference Room ____________ SYMBOLIC SYSTEMS FORUM The Interface between the Symbolic and the Subsymbolic David Rumelhart (der@psych.stanford.edu) Friday, 12 May, 3:15, 60:62N Abstract in next week's Calendar Looking for a Theory of Information Content Keith Devlin (devlin@csli.stanford.edu) Friday, 5 May, 3:15, 60:62N Abstract in last week's Calendar ____________ CSLI VISITOR CSLI welcomes its first Industrial Affiliate researcher, Ryo Ochitani, from Fujitsu Laboratories Ltd., Japan. Mr. Ochitani will be here until April 1990. For the past five years, he has helped develope the machine translation system ATLAS2 at Fujitsu Laboratories. During the last year, he and several other people started a new study on a quicker and easier system to classify sentences based on semantic view. As a start, they collected two million Japanese text samples, and classified several hundred sentences manually. Mr. Ochitani is interested in studying what kind of system would utilize this information most efficiently. If you wish to contact him, his email address is ochitani@csli.stanford.edu. ------------------------------ To: nl-kr@cs.rpi.edu Subject: Roger Ratcliff AI Seminar Reply-To: wsalter@BBN.COM Date: Thu, 27 Apr 89 18:20:52 EDT >From: wsalter@BBN.COM AI SEMINAR WEDS, MAY 3, 1:30 PM (NOTE ODD TIME) SECOND FLOOR LARGE CONFERENCE ROOM, BBN, 10 MOULTON ST. ROGER RATCLIFF, NORTHWESTERN UNIVERSITY CONNECTIONIST MODELS OF MEMORY AND FACTS ABOUT HUMAN MEMORY In this talk, connectionist models of memory are proposed and evaluated. The main class of models uses variants of the multilayer encoder model with the backpropagation delta rule for learning. The autoassociative model with the delta rule for learning is also evaluated in a much less exhaustive fashion. The models are applied to standard list learning procedures in which items are studied in a learning phase and then tested for retention. In contrast to most implementations of connectionist models, items or vectors are presented for learning one at a time (or in small groups of items within a rehearsal buffer) and an item is not trained further during the remainder of the list. This scheme mimics learning in many traditional memory experiments in which long lists of words are presented without repetition. This sequential learning scheme leads to two serious and central problems for the multilayer model. First, well learned information is forgotten rapidly as new information is learned. Second, discrimination between studied items and new items either decreases or is nonmonotonic as a function of the amount of rehearsal or number of learning trials each studied item receives. Both these results provide problems for the model because both predictions are inconsistent with large bodies of data from memory research. To address the first problem, manipulations of the network within the multilayer model were examined (for example holding some of the network weights constant and adding extra hidden units) but none of these significantly affected the forgetting functions. To address the second problem, several variants on the multilayer model were examined including learning as increments on a prelearned memory system, modifications only to a context portion of an item vector, and training an additional node to represent old items when turned on at test. None of these modifications produced adequate discrimination between studied and new items as a function of the amount of learning. The constraints demonstrated by these studies are important because they provide limitations on the scope and role of connectionist models in learning and forgetting in human memory and more generally in situations in which the whole set of information to be learned is not all available throughout learning. ------------------------------ End of NL-KR Digest *******************