Structured Language Modeling for Speech Recognition

TL;DR

This paper introduces a structured language model that incrementally builds hierarchical syntactic structures to improve speech recognition accuracy, outperforming traditional trigram models in perplexity and word error rate.

Contribution

It presents a novel structured language model that incorporates hierarchical syntax into speech recognition, enhancing performance over existing trigram models.

Findings

Improved perplexity (PPL) on WSJ corpus

Reduced word error rate (WER) compared to trigram models

Effective integration of hierarchical structure in language modeling

Abstract

A new language model for speech recognition is presented. The model develops hidden hierarchical syntactic-like structure incrementally and uses it to extract meaningful information from the word history, thus complementing the locality of currently used trigram models. The structured language model (SLM) and its performance in a two-pass speech recognizer --- lattice decoding --- are presented. Experiments on the WSJ corpus show an improvement in both perplexity (PPL) and word error rate (WER) over conventional trigram models.