Speaker Adaptation for Dialogue Act Classification
スポンサーリンク
概要
- 論文の詳細を見る
In this paper we investigate MAP adaptation to speakers for dialog act classification systems based on conditional random fields. MAP adaptation is done by assuming a Gaussian prior of the model-weights with mean equal to the weights of a baseline model. We did experiments on the ICSI meeting corpus and found that speaker adaptation gives significant improvements of the dialog act classification accuracy.
- 2011-07-14
著者
-
Koichi Shinoda
Tokyo Institute of Technology
-
Koichi Shinoda
Tokyo Insitute Of Technology
-
Johan Rohdin
Tokyo Insitute Of Technology
関連論文
- Inter-speaker weighted MAP adaptation for GMM-supervector speaker recognition
- Optimal use of trees in structural MAP adaptation for speaker verification
- Speaker Adaptation for Dialogue Act Classification
- Fusing deep speaker specific features and MFCC for robust speaker verification