Leveraging Cross-Utterance Context For ASR Decoding

TL;DR

This paper explores the use of long-context transformer language models during beam search decoding to incorporate cross-utterance information in speech recognition, showing improvements especially on long-format datasets.

Contribution

It introduces a method for integrating long-context transformer LMs into the first-pass decoding process for ASR, enhancing the utilization of cross-utterance context.

Findings

0.7% and 0.3% absolute WER reduction on AMI dev and test sets.

Improved performance with up to 500 tokens of prior context.

Less significant improvements (~0.1%) on Tedlium-1.

Abstract

While external language models (LMs) are often incorporated into the decoding stage of automated speech recognition systems, these models usually operate with limited context. Cross utterance information has been shown to be beneficial during second pass re-scoring, however this limits the hypothesis space based on the local information available to the first pass LM. In this work, we investigate the incorporation of long-context transformer LMs for cross-utterance decoding of acoustic models via beam search, and compare against results from n-best rescoring. Results demonstrate that beam search allows for an improved use of cross-utterance context. When evaluating on the long-format dataset AMI, results show a 0.7\% and 0.3\% absolute reduction on dev and test sets compared to the single-utterance setting, with improvements when including up to 500 tokens of prior context. Evaluations are also provided for Tedlium-1 with less significant improvements of around 0.1\% absolute.