Argumentation in lexicon and cognitive modelling

Abstract

Since the first studies on language, argumentation has been considered, with very few exceptions, as a phenomenon which had to be accounted for only after the meaning of the sentences under consideration had been extracted. Formal logic is among the best known tools generally used to extract meaning from sentences. As a confirmation of a very ancient analogy between meaning and knowledge, the same tool now prevails in knowledge acquisition and, of course, in knowledge representation. Formal logic is/spl minus/almost unquestionably/spl minus/considered as the fundamental structure of meaning and of knowledge. I question this apparently commonsensical position and propose a standpoint in which argumentation plays an essential role in knowledge management. The treatment of argumentative inference proposed, and the cognitive modelling that stems from it, have family resemblances with qualitative reasoning, to which it has, in the past, been compared, in some of their aspects. Several areas of artificial intelligence may benefit (and some already have) from this linguistic and cognitive re-consideration, from knowledge base semantic validation, to knowledge acquisition from text and to generation of explanations. >