Mode of access: World Wide Web.

System requirements: Adobe Acrobat Reader.

Part of: Synthesis digital library of engineering and computer science.

Series from website.

Includes bibliographical references (p. 137-148).

Preface -- 1. Introduction -- 1.1 Speech recognition and computation -- 1.2 Why WFST? -- 1.3 Purpose of this book -- 1.4 Book organization --

2. Brief overview of speech recognition -- 2.1 Statistical framework of speech recognition -- 2.2 Speech analysis -- 2.3 Acoustic model -- 2.3.1 Hidden Markov model -- 2.3.2 Computation of acoustic likelihood -- 2.3.3 Output probability distribution -- 2.4 Subword models and pronunciation lexicon -- 2.5 Context-dependent phone models -- 2.6 Language model -- 2.6.1 Finite-state grammar -- 2.6.2 N-gram model -- 2.6.3 Back-off smoothing -- 2.7 Decoder -- 2.7.1 Viterbi algorithm for continuous speech recognition -- 2.7.2 Time-synchronous Viterbi beam search -- 2.7.3 Practical techniques for LVCSR -- 2.7.4 Context-dependent phone search network -- 2.7.5 Lattice generation and N-best search --

3. Introduction to weighted finite-state transducers -- 3.1 Finite automata -- 3.2 Basic properties of finite automata -- 3.3 Semiring -- 3.4 Basic operations -- 3.5 Transducer composition -- 3.6 Optimization -- 3.6.1 Determinization -- 3.6.2 Weight pushing -- 3.6.3 Minimization -- 3.7 Epsilon removal --

4. Speech recognition by weighted finite-state transducers -- 4.1 Overview of WFST-based speech recognition -- 4.2 Construction of component WFSTs -- 4.2.1 Acoustic models -- 4.2.2 Phone context dependency -- 4.2.3 Pronunciation lexicon -- 4.2.4 Language models -- 4.3 Composition and optimization -- 4.4 Decoding algorithm using a single WFST -- 4.5 Decoding performance --

5. Dynamic decoders with on-the-fly WFST operations -- 5.1 Problems in the native WFST approach -- 5.2 On-the-fly composition and optimization -- 5.3 Known problems of on-the-fly composition approach -- 5.4 Look-ahead composition -- 5.4.1 How to obtain prospective output labels -- 5.4.2 Basic principle of look-ahead composition -- 5.4.3 Realization of look-ahead composition using a filter transducer -- 5.4.4 Look-ahead composition with weight pushing -- 5.4.5 Generalized composition -- 5.4.6 Interval representation of label sets -- 5.5 On-the-fly rescoring approach -- 5.5.1 Construction of component WFSTs for on-the-fly rescoring -- 5.5.2 Concept -- 5.5.3 Algorithm -- 5.5.4 Approximation in decoding -- 5.5.5 Comparison with look-ahead composition --

6. Summary and perspective -- 6.1 Realization of advanced speech recognition techniques using WFSTs -- 6.1.1 WFSTs for extended language models -- 6.1.2 Dynamic grammars based on WFSTs -- 6.1.3 Wide-context-dependent HMMs -- 6.1.4 Extension of WFSTs for multi-modal inputs -- 6.1.5 Use of WFSTs for learning -- 6.2 Integration of speech and language processing -- 6.3 Other speech applications using WFSTs -- 6.4 Conclusion --

Bibliography -- Authors' biographies.

Abstract freely available; full-text restricted to subscribers or individual document purchasers.

Compendex

INSPEC

Google scholar

Google book search

This book introduces the theory, algorithms, and implementation techniques for efficient decoding in speech recognition mainly focusing on the Weighted Finite-State Transducer (WFST) approach. The decoding process for speech recognition is viewed as a search problem whose goal is to find a sequence of words that best matches an input speech signal. Since this process becomes computationally more expensive as the system vocabulary size increases, research has long been devoted to reducing the computational cost. Recently, the WFST approach has become an important state-of-the-art speech recognition technology, because it offers improved decoding speed with fewer recognition errors compared with conventional methods. However, it is not easy to understand all the algorithms used in this framework, and they are still in a black box for many people. In this book, we review the WFST approach and aim to provide comprehensive interpretations of WFST operations and decoding algorithms to help anyone who wants to understand, develop, and study WFST-based speech recognizers. We also mention recent advances in this framework and its applications to spoken language processing.

Also available in print.

Title from PDF t.p. (viewed on January 18, 2013).