We have developed a knowledge base which describes the semantic co-occurrence between words and the inference mechanism on it. Using this knowledge base, it is possible to predict the next words likely to be uttered in the context. Also, this knowledge base is automatically created from ATR's linguistic database and can easily be constructed for a large vocabulary in a specific domain. This paper describes the structure of the knowledge base and the inference method.
Bibliographic reference. Inoue, Naomi / Morimoto, Tsuyoshi / Ogura, Kentarou (1989): "A linguistic knowledge base for applying semantic information to a speech understanding system", In EUROSPEECH-1989, 1194-1197.