Paper Info
Title
A comparative study on methods of Weighted language model training for reranking lvcsr N-best hypotheses
Abstract
This paper focuses on discriminative n-gram language models for a large vocabulary speech recognition task. Specifically we compare three training methods, Reranking Boosting (ReBst), Minimum Error Rate Training (MERT) and the Weighted Global Log-Linear Model (W-GCLM). They have a mechanism for handling sample weights, which are useful for providing an accurate model and work as impact factors of hypotheses for training. W-GCLM is proposed in this paper. We discuss the relationship between the three methods by comparing their loss functions. We also compare them experimentally by reranking N-best hypotheses under several conditions. We show that MERT and W-GCLM are different types of expansion of ReBst and have different respective advantages. Our experimental results reveal that W-GCLM outperforms ReBst and whether MERT or W-GCLM is superior depends on the training and test conditions.
Year
DOI
Venue
2010
10.1109/ICASSP.2010.5495028
Acoustics Speech and Signal Processing
Keywords
Field
DocType
speech recognition,N-best hypotheses reranking,discriminative n-gram language model,loss function,minimum error rate training,reranking boosting,vocabulary speech recognition task,weighted global log-linear model,weighted language model training,Discriminative LM,Error Correction,MERT,Reranking Boost,Weighted GCLM
Pattern recognition,Computer science,Word error rate,Natural language,Boosting (machine learning),Artificial intelligence,Estimation theory,Hidden Markov model,Vocabulary,Discriminative model,Language model,Machine learning
Conference
ISSN
ISBN
Citations 
1520-6149 E-ISBN : 978-1-4244-4296-6
978-1-4244-4296-6
12
PageRank 
References 
Authors
0.60
7
3
Name
Order
Citations
PageRank
Takanobu Oba15312.09
Takaaki Hori240845.58
Atsushi Nakamura35713.11