Computational Approaches to the Pragmatics Problem

Unlike many aspects of human language, pragmatics involves a systematic many-to-many mapping between form and meaning. This renders the computational problems of encoding and decoding meaning especially challenging, both for humans in normal conversation and for artificial dialogue systems that need to understand their users' input. A particularly striking example of this difficulty is the recognition of speech act or dialogue act types. In this review, we discuss why this is a problem and why its solution is potentially relevant both for our understanding of human interaction and for the implementation of artificial systems. We examine some of the theoretical and practical attempts that have been made to overcome this problem and consider how the field might develop in the near future.