Model Shrinkage for Discriminative Language Models

Takanobu OBA  Takaaki HORI  Atsushi NAKAMURA  Akinori ITO  

IEICE TRANSACTIONS on Information and Systems   Vol.E95-D   No.5   pp.1465-1474
Publication Date: 2012/05/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E95.D.1465
Print ISSN: 0916-8532
Type of Manuscript: PAPER
Category: Speech and Hearing
model shrinkage problem,  discriminative language model,  linear model,  feature selection,  

Full Text: PDF(466.1KB)
>>Buy this Article

This paper describes a technique for overcoming the model shrinkage problem in automatic speech recognition (ASR), which allows application developers and users to control the model size with less degradation of accuracy. Recently, models for ASR systems tend to be large and this can constitute a bottleneck for developers and users without special knowledge of ASR with respect to introducing the ASR function. Specifically, discriminative language models (DLMs) are usually designed in a high-dimensional parameter space, although DLMs have gained increasing attention as an approach for improving recognition accuracy. Our proposed method can be applied to linear models including DLMs, in which the score of an input sample is given by the inner product of its features and the model parameters, but our proposed method can shrink models in an easy computation by obtaining simple statistics, which are square sums of feature values appearing in a data set. Our experimental results show that our proposed method can shrink a DLM with little degradation in accuracy and perform properly whether or not the data for obtaining the statistics are the same as the data for training the model.