Title
Two-step generation of variable-word-length language model integrating local and global constraints
Abstract
This paper proposes two-step generation of a variable-length class-based language model that integrates local and global constraints. In the first-step, an initial class set is recursively designed using local constraints. Word elements for each class are determined using Kullback divergence and total entropy. In the second step, the word classes are recursively and words are iteratively recreated, by grouping consecutive words to generate longer units and by splitting the initial classes into finer classes. These operations in the second step are carried out selectively, taking into account local and global constraints on the basis of a minimum entropy criterion. Experiments showed that the perplexity of the proposed initial class set is superior to that of the conventional part-of-speech class, and the perplexity of the variable-word-length model consequently becomes lower. Furthermore, this two-step model generation approach greatly reduces the training time
Year
DOI
Venue
1998
10.1109/ICASSP.1998.675360
Acoustics, Speech and Signal Processing, 1998. Proceedings of the 1998 IEEE International Conference
Keywords
Field
DocType
grammars,iterative methods,minimum entropy methods,natural languages,speech processing,speech recognition,Kullback divergence,class-based language model,experiments,global constraints,iterative word-class,large vocabulary continuous speech recognition,local constraints,minimum entropy criterion,part-of-speech class,perplexity,total entropy,training time reduction,two-step model generation,variable-word-length language model
Rule-based machine translation,Speech processing,Perplexity,Pattern recognition,Iterative method,Computer science,Natural language,Artificial intelligence,Vocabulary,Language model,Recursion
Conference
Volume
ISSN
ISBN
2
1520-6149
0-7803-4428-6
Citations 
PageRank 
References 
0
0.34
6
Authors
2
Name
Order
Citations
PageRank
Shoichi Matsunaga116436.02
Shigeki Sagayama21217137.97