Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data

Abstract: 

Bidirectional long short-term memory (BLSTM) recurrent neural network (RNN) has achieved state-of-the-art performance in many sequence processing problems given its capability in capturing contextual information. However, for languages with limited amount of training data, it is still difficult to obtain a high quality BLSTM model for emphasis detection, the aim of which is to recognize the emphasized speech segments from natural speech. To address this problem, in this paper, we propose a multilingual BLSTM (MTL-BLSTM) model where the hidden layers are shared across different languages while the softmax output layer is language-dependent. The MTL-BLSTM can learn cross-lingual knowledge and transfer this knowledge to both languages to improve the emphasis detection performance. Experimental results demonstrate our method can outperform the comparison methods over 2-15.6% and 2.9-15.4% on the English corpus and Mandarin corpus in terms of relative F1-measure, respectively.

up
0 users have voted:

Paper Details

Authors:
Yishuang Ning, Zhiyong Wu, Runnan Li, Jia Jia, Mingxing Xu, Helen Meng, Lianhong Cai
Submitted On:
4 March 2017 - 10:26am
Short Link:
Type:
Poster
Event:
Presenter's Name:
Yishuang Ning
Paper Code:
2781
Document Year:
2017
Cite

Document Files

ICASSP2017-Poster presentation-horizontal-v2-nys.pptx

(206 downloads)

Subscribe

[1] Yishuang Ning, Zhiyong Wu, Runnan Li, Jia Jia, Mingxing Xu, Helen Meng, Lianhong Cai, "Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data", IEEE SigPort, 2017. [Online]. Available: http://sigport.org/1626. Accessed: Feb. 21, 2019.
@article{1626-17,
url = {http://sigport.org/1626},
author = {Yishuang Ning; Zhiyong Wu; Runnan Li; Jia Jia; Mingxing Xu; Helen Meng; Lianhong Cai },
publisher = {IEEE SigPort},
title = {Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data},
year = {2017} }
TY - EJOUR
T1 - Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data
AU - Yishuang Ning; Zhiyong Wu; Runnan Li; Jia Jia; Mingxing Xu; Helen Meng; Lianhong Cai
PY - 2017
PB - IEEE SigPort
UR - http://sigport.org/1626
ER -
Yishuang Ning, Zhiyong Wu, Runnan Li, Jia Jia, Mingxing Xu, Helen Meng, Lianhong Cai. (2017). Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data. IEEE SigPort. http://sigport.org/1626
Yishuang Ning, Zhiyong Wu, Runnan Li, Jia Jia, Mingxing Xu, Helen Meng, Lianhong Cai, 2017. Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data. Available at: http://sigport.org/1626.
Yishuang Ning, Zhiyong Wu, Runnan Li, Jia Jia, Mingxing Xu, Helen Meng, Lianhong Cai. (2017). "Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data." Web.
1. Yishuang Ning, Zhiyong Wu, Runnan Li, Jia Jia, Mingxing Xu, Helen Meng, Lianhong Cai. Learning Cross-lingual Knowledge with Multilingual BLSTM for Emphasis Detection with Limited Training Data [Internet]. IEEE SigPort; 2017. Available from : http://sigport.org/1626