Abstract
In artificial intelligence, human understanding of text is modeled by programs which construct representations. A critical question concerns determining the form of the representation (hence the form of the understanding). The notion of coherence emphasizes the connectivity of items in the understanding/representation. Coherence representations either tie together the words of the text, the discourse units of the text, or the concepts the text invokes. The coherence viewpoint is exemplified by the works of Halliday & Hasan (1976) on cohesion, Lockman & Klappholz (1980) on contextual reference resolution, Hobbs (1979) on rhetorical coherence, and the semantic network-derived text representation techniques of Norvig (1989), Kintsch (1988), and Charniak (1986, 1983).