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

facebooktwittermailshare

LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION

Abstract: 

In this paper, we explore several new schemes to train a seq2seq model to integrate a pre-trained language model (LM). Our proposed fusion methods focus on the memory cell state and the hidden state in the seq2seq decoder long short-term memory (LSTM), and the memory cell state is updated by the LM unlike the prior studies. This means the memory retained by the main seq2seq would be adjusted by the external LM. These fusion methods have several variants depending on the architecture of this memory cell update and the use of memory cell and hidden states which directly affects the final label inference. We performed the experiments to show the effectiveness of the proposed methods in a mono-lingual ASR setup on the Librispeech corpus and in a transfer learning setup from a multilingual ASR (MLASR) base model to a low-resourced language. In Librispeech, our best model improved WER by 3.7%, 2.4% for test clean, test other relatively to the shallow fusion baseline, with multilevel decoding. In transfer learning from an MLASR base model to the IARPA Babel Swahili model, the best scheme improved the transferred model on eval set by 9.9%, 9.8% in CER, WER relatively to the 2-stage transfer baseline.

up
0 users have voted:

Paper Details

Authors:
Jaejin Cho, Shinji Watanabe, Takaaki Hori, Murali Karthick Baskar, Hirofumi Inaguma, Jesus Villalba, Najim Dehak
Submitted On:
10 May 2019 - 4:50pm
Short Link:
Type:
Poster
Event:
Presenter's Name:
Jaejin Cho
Paper Code:
SLP-P5.9
Document Year:
2019
Cite

Document Files

Poster

(12)

Subscribe

[1] Jaejin Cho, Shinji Watanabe, Takaaki Hori, Murali Karthick Baskar, Hirofumi Inaguma, Jesus Villalba, Najim Dehak, "LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION", IEEE SigPort, 2019. [Online]. Available: http://sigport.org/4405. Accessed: Jun. 18, 2019.
@article{4405-19,
url = {http://sigport.org/4405},
author = {Jaejin Cho; Shinji Watanabe; Takaaki Hori; Murali Karthick Baskar; Hirofumi Inaguma; Jesus Villalba; Najim Dehak },
publisher = {IEEE SigPort},
title = {LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION},
year = {2019} }
TY - EJOUR
T1 - LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION
AU - Jaejin Cho; Shinji Watanabe; Takaaki Hori; Murali Karthick Baskar; Hirofumi Inaguma; Jesus Villalba; Najim Dehak
PY - 2019
PB - IEEE SigPort
UR - http://sigport.org/4405
ER -
Jaejin Cho, Shinji Watanabe, Takaaki Hori, Murali Karthick Baskar, Hirofumi Inaguma, Jesus Villalba, Najim Dehak. (2019). LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION. IEEE SigPort. http://sigport.org/4405
Jaejin Cho, Shinji Watanabe, Takaaki Hori, Murali Karthick Baskar, Hirofumi Inaguma, Jesus Villalba, Najim Dehak, 2019. LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION. Available at: http://sigport.org/4405.
Jaejin Cho, Shinji Watanabe, Takaaki Hori, Murali Karthick Baskar, Hirofumi Inaguma, Jesus Villalba, Najim Dehak. (2019). "LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION." Web.
1. Jaejin Cho, Shinji Watanabe, Takaaki Hori, Murali Karthick Baskar, Hirofumi Inaguma, Jesus Villalba, Najim Dehak. LANGUAGE MODEL INTEGRATION BASED ON MEMORY CONTROL FOR SEQUENCE TO SEQUENCE SPEECH RECOGNITION [Internet]. IEEE SigPort; 2019. Available from : http://sigport.org/4405