Lexical Knowledge for Natural Language Generation

In this paper, we examine how lexical knowledge has been modelled and used in generation systems of the past, and, looking to the future, what will be required of the lexicon to build versatile and expressive generators. Three central issues discussed are the following: the relationship between words and text structure, the connection between concepts and words, and the influence of context on word use and combination. We find that for a generator to be expressive, it must have knowledge of phrasal items, especially of collocations. We also find that the many available options for expressing a concept can be distinguished from another by pragmatic factors such as connotation, implicature, and style. Most importantly, this phrasal and pragmatic knowledge should be explicit and functional; it must prescribe how words are used, chosen, and combined. Copyright c fl 1995 by Philip Glenny Edmonds. "Every concept that can ever be needed, will be expressed by exactly one word, with its meani..