Integrated Processing in Planning and Understanding.

Programs that plan and understand must make many decisions about which paths of inquiry are likely to prove fruitful. In order to make such decisions rationally, and hence avoid the need for backtracking that inevitably results if they are made arbitrarily, relevant contextual information must be brought to bear. An integrated model of planning or understanding is one that attempts to take such contextual information into account as early as possible. An integrated model of understanding must take the understanders goals and hypotheses into account in making decisions about how to interpret an input. The relationship between syntax and semantics in language understanding is analyzed from such an integrated point of view. Next, the problems of lexical ambiguity and vagueness are addressed, previous attempts to solve these problems are analyzed, and their shortcomings are used to motivate requirements for a more complete solution. Finally, an integrated approach to inference in explanation-based understanding is presented. Keywords Artificial intelligence, Language understanding, Opportunistic planning, Explanatory inference, Freudian slips.