| If you hear the sentences 'there are only five birds in this tree' and 'there are at most five parrots on the branches of this tree', you will immediately know that the meaning of the latter sentence is contained in the meaning of the former. This informational relation is an example of what formal semanticists call *entailment*. Entailment relations exhibit a complex interplay of linguistic and extra-linguistic principles involving logic, meaning and use. How are humans able to establish such complex relations between utterances? Answering this question is of utmost importance for exploring the human abilities in language and reasoning. Traditional formal semantics of natural language have no problem analyzing some extremely simple entailments with `logical' expressions like `five', `only' and `at most'. In the above example, however, meaning relations between content words like `birds'-`parrots', `in'-`on', and `tree'-`branches' should also be accounted for. Standard logical approaches to language are of little help when analyzing such relations, which are indispensable for analyzing many entailments. Consequently, despite their centrality in language and reasoning, entailment relations have remained an immense puzzle. In a unique integration of principles from formal semantics, experimental linguistics and artificial intelligence, this program will develop a model of entailment that incorporates the intricacies of content word meanings. The leading hypothesis is that logical operators in formal semantics are sensitive to fuzzy but mathematically well-defined elements in the common sense meaning of content words. This hypothesis serves as a central building block in the new framework of Content Sensitive Formal Semantics. With its radical reorganization of formal semantic theory, this framework is expected to constitute a breakthrough in our understanding of linguistic reasoning. |
