Towards a computational model of humor

The phenomena of humor recognition seems to occur, in part, at cognitive levels that are below conscious awareness. Thus, the details of how humorous statements are recognized are not open to introspection or simple investigatory techniques. In this thesis, an information processing model of humor appreciation was further specified by integrating it with three concurrent models of natural language understanding. Then detailed processes for recognizing a written form of humor (puns) were couched in terms of syntactic and semantic processes, and the model was programmed in LISP for the purpose of testing its plausibility. It was concluded that the program can recognize simple forms of potentially humorous sentences, as well as resolve the ambiguity of many words m ordinary sentences. The program illuminates several of the detailed processes involved in the recognition of potentially humorous statements and in the resolution of ambiguous words, and provides a new approach for future research in the area of humor appreciation.