Speech Recognition by Composition of Weighted Finite Automata

TL;DR

This paper introduces a unified framework using weighted finite automata and transducers for speech recognition, enabling efficient combination and optimization of various information sources like language models, dictionaries, and acoustic data.

Contribution

It presents a novel, general composition algorithm that integrates multiple information sources in speech recognition within a unified automata-based framework.

Findings

Unified representation of recognition components

Efficient algorithms for source combination and optimization

Single composition algorithm for static and dynamic integration

Abstract

We present a general framework based on weighted finite automata and weighted finite-state transducers for describing and implementing speech recognizers. The framework allows us to represent uniformly the information sources and data structures used in recognition, including context-dependent units, pronunciation dictionaries, language models and lattices. Furthermore, general but efficient algorithms can used for combining information sources in actual recognizers and for optimizing their application. In particular, a single composition algorithm is used both to combine in advance information sources such as language models and dictionaries, and to combine acoustic observations and information sources dynamically during recognition.