The computation, representation and processing of semantic information is an important task in the development of natural language systems. There has been a noticeable convergence of ideas in recent years, centering on the use of contextual information and dynamic aspects of utterances on the one hand, and the need for structured information, which can be differentiated according to a specific domain of application, on the other hand. Several formalisms are available, which approach problems related to these ideas with different techniques and from different angles. The colloquium has contributed to clarifying the issues surrounding the use of these formalisms in semantics research both in theoretical semantics and in artificial intelligence.

Every day of the colloquium was devoted to one main theme:

• 23.02.93: Semantics Formalisms
• 24.02.93: Syntax-Interface / Underspecified Representation
• 25.02.93: Knowledge Processing and Semantics Formalisms
• 26.02.93: Non-monotonic Reasoning and Semantics Formalisms

Many well-known experts in semantics, logic, computational linguistics and computer science took part in the seminar, all of whom have made important contributions to research in these fields.

The interdisciplinary orientation of the seminar has made it possible

• to enhance the state of information in the relevant disciplines;
• to provide an overview of existing implementations and their power;
• to illuminate the theoretic status of different decisions in implementation;
• to improve our estimates of the practical potential of theoretical semantics fomialisms.

Great emphasis was placed on the exchange of ideas between the different fields and discussions between the different approaches. Furthermore, a special working group examined a catalog of criteria for evaluating the different existing approaches. This work, and subsequent plenary discussion, isolated several parameters - pertaining to the realization state, the theoretical intentions, and the coverage relative to which the systems differ to a great extent. Further careful planning will be necessary to arrive at meaningful comparative evaluations. A practical consequence of this has been an initiative to create a special interest group with the Association of Computational Linguistics on the topic of computational semantics. It was also agreed to cooperate on creating a database of representative examples of relevant natural language phenomena.

The participants felt that the colloquium was a very successful and stimulating event. In light of the importance of such research exchange, the next meeting is projected to take place in 1995. It is to be concerned specifically with the evaluation of formalisms which have been implemented and are used in applications.

Creative Commons BY 3.0 Unported license