Speech Dialogue System with the Function of Associating Words
Grant-in-Aid for Scientific Research (C).
|Research Institution||Yamanashi University|
関口 芳広 山梨大学, 工学, 助教授
SEKIGUCHI Yoshihiro Yamanashi university, Department of Engineering, associated professor, 工学部, 助教授 (70020493)
SHIGENAGA Minoru Chukyo university, School of Computer and Cognitive Sciences, professor, 情報科学科, 教授 (20020282)
ARIIZUMI Hitoshi Yamanashi university, Department of Engineering, assistant professor, 工学部, 講師 (80020436)
KARASAWA Hiroshi Yamanashi university, Department of Engineering, associated professor, 工学部, 助教授 (90177618)
|Project Fiscal Year
1992 – 1993
Completed(Fiscal Year 1993)
|Budget Amount *help
¥1,500,000 (Direct Cost : ¥1,500,000)
Fiscal Year 1993 : ¥600,000 (Direct Cost : ¥600,000)
Fiscal Year 1992 : ¥900,000 (Direct Cost : ¥900,000)
|Keywords||Speech Dialogue System / Associative information / Speech recognition / Speech synthesis by rule / Dialogue understanding / Dialogue contorol / 音声対話 / 連想情報 / 連続音声認識 / 規則音声合成 / 対話理解 / 対話制御 / 音声対話システム / 連結音声認識 / 連想 / 音声合成|
We have proposed the use of associative information to a speech dialogue system in addition to syntax and semantic information. This research is concerned with the following.
(1) Human acts of associating words with given words are investigated, and an associative word dictionary has been constructed for speech dialogue system.
(2) A procedure which predicts following words by using associative information has been proposed.
(3) A method which calculates matching scores between input speech and reference patterns by using both acoustic and associative information has been proposed.
(4) Continous speech recognition rates has been improved by using associative information.
(5) It has been ascertained that for continuous speech recognition the best ratio mixing a score due to associative information with one due to acoustic information is about 1 : 3.
(6) We have constructed a speech dialogue system for the guidance of city hall, it has the function of associating words. This system has been able to deal with about 80% of the dialogue correctly.
We have also developed the following elemental techniques for a high level verbal communication between a man and a machine.
(1) A method for creating a system's goal named Social Relation Analisys has been developed. A system's goal is here derived from user's requirement made by both direct and indirect speech act.
(2) Polite expression is very important but complicated in Japanese. This type of expression controls the mental distance between a user and a system. We have implemented the experimental system to define the necesity of polite expression by estimating the three kind of social relations.
(3) To control the coherence of topic is neccesary for an efficient dialogue. We make a dialogue tree from dialogue sentences, and calculate the ratio of the coherence of topic with thesaurus. We have another five kind of reseach results but eliminate them here.
Research Output (16results)