Abstract

In this paper, we introduce a method of generating a prosodic tree structure from the F0 contour of an utterance in order to analyze the information expressed by prosody in Japanese spontaneous dialogue. The connection rate, which means the strength of the relationship between two prosodic units, is de ned. By repeatedly combining the two adjacent prosodic units where the rate is high, the tree structure is gradually generated. To determine the parameters objectively, we applied the principal component analysis to 32 dialogues from the Chiba Map Task Dialogue Corpus. Then we applied our method to one dialogue. The results suggested that the prosodic tree based on the rst principal component was concerned with the information telling what the speaker wanted to do next and that the prosodic tree based on the second principal component represented the syntactic and grammatical structure.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call