What Must Computer Understanding Systems Understand?

This paper attempts to characterize some of the knowledge that people have of their language's semantic structure, in order to determine what a computer has to know to perform the operations of searching for facts and answering questions. Necessarily, this sort of semantic analysis is done in connection with syntactic structure. The meaningful elements of English are divided into sets of words and sets of sentences. Their general properties are examined as well as the general relations between elements. The paper concludes with a discussion of the controversy between generative semanticists and interpretivists over syntactic and semantic structure. The work of logicians on artificial language is cited as being helpful in explanations of syntactic and semantic structure in natural languages. (AM)