Hidekazu Arita

Predicting noun-phrase surface forms using contextual information

We propose a context-sensitive method to predict noun-phrases in the next utterance of a telephone inquiry dialogue. First, information about the utterance type and the discourse entities of the next utterance is grasped using a dialogue interpretation model. Second, a domain-dependent knowledge base for noun-phrase usage is developed, focusing on the dialogue situations in context. Finally, we propose a strategy to make a set of the appropriate expressions in the next utterance, using the information and the knowledge base. This set of expressions is used to select the correct candidate from the speech recognition output. This paper examines some of the processes creating sets of polite expressions, deictic expressions, and compound noun phrases, which are common in telephone inquiry dialogue.