aanttila@cc.helsinki.fi (05/03/91)
*********************************************************************** SYMBOL MANIPULATION, CONNECTIONISM AND THE SEMANTICS OF NATURAL LANGUAGE November 14-15, 1991 University of Helsinki Finland *********************************************************************** A two-day workshop to be held at the University of Helsinki, Finland. ORGANISERS The Linguistic Association of Finland (SKY) The Finnish Artificial Intelligence Society (STeS) The Philosophical Society of Finland (SFY) _______________________________________________________________________ INVITED SPEAKERS Graeme Hirst, Universities of Toronto and Rochester Ronald W. Langacker, University of California, San Diego Keith Stenning, University of Edinburgh, Centre for Cognitive Science _______________________________________________________________________ BACKGROUND In theoretical linguistics, symbolic representation has a long and venerable tradition. Since de Saussure, natural language has been seen as an aggregate of discrete symbols, and, by consequence, the linguistic metalanguage (theories, descriptions) has been thoroughly symbolic as well. This is certainly true of the Chomskian paradigm as well as most of its alternatives (GPSG, HPSG, LFG, CG,...). The AI community has largely adopted a symbol manipulation paradigm. The vast majority of systems aiming at natural language understanding are based on rule-like formalisms. The results have been only partly satisfactory. Problems concerning the impreciseness and the ambiguity of natural language have often led to dead ends, lexical semantics has hardly been touched upon. Recently, artificial neural networks have been proposed as one possible solution. In philosophy, many of the currently relevant problems of AI have a long history. Concepts such as intelligence, knowledge, sign and meaning have long been a central concern particularly in epistemology and the philosophy of mind. An understanding of the history of ideas will give new insight into the nature of the basic problems in AI. _______________________________________________________________________ TOPICS The aim of the workshop is to relate the philosophical and linguistic aspects of natural language semantics with the methodological alternatives of symbolic representation and reasoning and its connectionist alternative, artificial neural networks. The workshop will focus on the following topics: * Do cognitive activities require a language-like representational medium? * Is language of thought needed to explain natural language understanding? * How can the interface between language and reality be modelled? * What is the relation between natural language and logic? Are the phenomena understated because of the methods used? * What is the relationship between syntax and semantics? How should one define the borderline between semantics and pragmatics? * What are the merits of traditional natural language processing systems? What are the main deficiencies of present machine translation systems? * What improvements can be made following the symbolist paradigm by simply creating larger knowledge bases? Are there more profound methodological problems? * What are the results of connectionism so far concerning natural language semantics? What results can one expect by the end of this decade? * What are the counterparts of connectionism in philosophy and linguistics? * Does connectionism give rise to a re-evaluation of some central problems in Western philosophy? _______________________________________________________________________ CONTACT INFORMATION For more information, please contact Arto Anttila Secretary of the Linguistic Association of Finland Research Unit for Computational Linguistics University of Helsinki Hallituskatu 11-13 SF-00100 HELSINKI Finland e-mail: aanttila@finuh.bitnet tel: 358-0-191 3500 _______________________________________________________________________