Generation and Pruning of Pronunciation Variants to Improve ASR Accuracy

TL;DR

This paper introduces a data-driven method for generating and pruning pronunciation variants to enhance automatic speech recognition accuracy, especially for names, by updating pronunciation dictionaries without harming recognition of similar words.

Contribution

It presents an efficient technique that automatically learns acceptable pronunciations and updates lexicons, significantly reducing error rates in name recognition tasks.

Findings

Reduced name recognition error rate by 42%.

Generalizes well across datasets of various sizes.

Improves recognition accuracy without affecting similar words.

Abstract

Speech recognition, especially name recognition, is widely used in phone services such as company directory dialers, stock quote providers or location finders. It is usually challenging due to pronunciation variations. This paper proposes an efficient and robust data-driven technique which automatically learns acceptable word pronunciations and updates the pronunciation dictionary to build a better lexicon without affecting recognition of other words similar to the target word. It generalizes well on datasets with various sizes, and reduces the error rate on a database with 13000+ human names by 42%, compared to a baseline with regular dictionaries already covering canonical pronunciations of 97%+ words in names, plus a well-trained spelling-to-pronunciation (STP) engine.