N-gram Boosting: Improving Contextual Biasing with Normalized N-gram Targets

TL;DR

This paper introduces N-gram Boosting, a method that enhances speech-to-text accuracy for proper names and technical terms by boosting normalized n-gram targets, significantly improving recognition rates especially for complex tokens.

Contribution

It proposes a novel two-step boosting mechanism for normalized n-grams, addressing missing hits and over-boosting issues in keyword recognition tasks.

Findings

26% relative improvement on in-domain dataset

2% improvement on LibriSpeech

Effective for non-standard and non-alphabetic targets

Abstract

Accurate transcription of proper names and technical terms is particularly important in speech-to-text applications for business conversations. These words, which are essential to understanding the conversation, are often rare and therefore likely to be under-represented in text and audio training data, creating a significant challenge in this domain. We present a two-step keyword boosting mechanism that successfully works on normalized unigrams and n-grams rather than just single tokens, which eliminates missing hits issues with boosting raw targets. In addition, we show how adjusting the boosting weight logic avoids over-boosting multi-token keywords. This improves our keyword recognition rate by 26% relative on our proprietary in-domain dataset and 2% on LibriSpeech. This method is particularly useful on targets that involve non-alphabetic characters or have non-standard pronunciations.