Of all the research areas in natural-language processing, the problem of sentence analysis has received the greatest attention in terms of computer implementations. Because sentence analysis is the first step for all other memory manipulations, it seems quite reasonable to worry about it first. After all, any natural-language processing program that cannot input natural language looks at least a little suspect right from the start. But, if sentence analysis qualifies as a topic in natural-language processing, it is a very broad one. In fact, many researchers have come to the conclusion that sentence analysis cannot be reasonably attacked without simultaneously addressing problems of knowledge representation, inference, long-term memory organization, and the social psychology of linguistic interactions. This holistic perspective brings new meaning to Wittgenstein's old claim that understanding a sentence is synonymous with understanding a language, but it does not make organizing a technical anthology any easier.
So, at the risk of offending any hard-time advocates of fully integrated systems, this section addresses the issue of sentence analysis as it is practiced by existing computer programs. Although sentence analysis per se is a perfectly general notion, most existing systems will practice sentence analysis as it is needed for a specific task orientation. Our authors discuss a number of specific task orientations, in-