Estimation of Dialogue States in Database Search Spoken Dialogue System
-
- Ryota Nishimura
- Department of Electrical Engineering and Computer Science, Graduate School of Engineering, Nagoya University, Japan.
-
- Kazunori Komatani
- Department of Electrical Engineering and Computer Science, Graduate School of Engineering, Nagoya University, Japan.
Bibliographic Information
- Other Title
-
- データベース検索音声対話システムにおける対話状態の推定
Search this article
Abstract
We describe an estimation model of dialogue states in spoken dialogue systems for the database search task. We model dialogues in the database search task as consisting of two states: "specifying retrieval conditions (search)" and "requesting detailed information about specific entries (info.)". The two states are predicted by a logistic regression classifier based on features obtained from the dialog. We developed a spoken dialogue system for the restaurant database search task and collected dialogue data from seven participants. The experimental result showed that the estimation accuracy was 97.4%. We investigated which features contributed to the estimation of the states.
Journal
-
- IPSJ SIG Notes
-
IPSJ SIG Notes 2012 (20), 1-7, 2012-01-27
Information Processing Society of Japan (IPSJ)
