Integrating Syntagmatic Information in a Dictionary for Computer Speech Applications

Conventional dictionaries, albeit they often comprise an impressive amount o f para­ digmatic information on various aspects o f linguistic description, usually pay only little attention to the representation o f syntagmatic information. Admittedly, apart hrom spelling conventions and rules o f inflectional agreement, the co-occurrence o f indivi­ dual lexical items will not normally change the orthographic shape o f a word when it appears in written text. In spoken language, however, the phonetic realization o f words is heavily influenced by context and may change dramatically in a variety o f ways, including segmental as well as prosodic features. These changes need to be taken into account in both computer speech synthesis and automatic qreech recognition. In this paper, dierefore, we argue fathe inclusion of syntagmatic information in dictionaries which are developed for the special purpose o f spdcen language processing in computer speech applications. Two kinds of syntagmatic information will be considered in more detail: Case Frames and Collocations.