8th European Conference on Speech Communication and Technology

Geneva, Switzerland
September 1-4, 2003


Prediction of Sentence Importance for Speech Summarization Using Prosodic Parameters

Akira Inoue, Takayoshi Mikami, Yoichi Yamashita

Ritsumeikan University, Japan

Recent improvements in computer systems are increasing the amount of accessible speech data. Since speech media is not appropriate for quick scanning, the development of automatic summarization of lecture or meeting speech is expected. Spoken messages contain non-linguistic information, which is mainly expressed by prosody, while written text conveys only linguistic information. There are possibilities that the prosodic information can improve the quality of speech summarization. This paper describes a technique of using prosodic parameters as well as linguistic information to identify important sentences for speech summarization. Several prosodic parameters about F0, power and duration are extracted for each sentence in lecture speech. Importance of the sentence is predicted by the prosodic parameters and the linguistic information. We also tried to combine the prosodic parameters and the linguistic information by multiple regression analysis. Proposed methods are evaluated both on the correlation between the predicted scores of sentence importance and the preference scores by subjects and on the accuracy of extraction of important sentences. By combination of the prosodic parameters improves the quality of speech summarization.

Full Paper

Bibliographic reference.  Inoue, Akira / Mikami, Takayoshi / Yamashita, Yoichi (2003): "Prediction of sentence importance for speech summarization using prosodic parameters", In EUROSPEECH-2003, 1193-1196.