HypR: A comprehensive study for ASR hypothesis revising with a reference corpus

TL;DR

This paper introduces HypR, a comprehensive dataset and benchmark for ASR hypothesis revising, facilitating fair comparison of error correction and reranking methods to improve speech recognition accuracy.

Contribution

It provides the first unified dataset with multiple corpora and hypotheses, along with implemented baseline methods, to standardize evaluation in ASR hypothesis revising research.

Findings

Compared various revising methods on the HypR dataset

Demonstrated the effectiveness of different error correction techniques

Established a benchmark for future ASR hypothesis revising studies

Abstract

With the development of deep learning, automatic speech recognition (ASR) has made significant progress. To further enhance the performance of ASR, revising recognition results is one of the lightweight but efficient manners. Various methods can be roughly classified into N-best reranking modeling and error correction modeling. The former aims to select the hypothesis with the lowest error rate from a set of candidates generated by ASR for a given input speech. The latter focuses on detecting recognition errors in a given hypothesis and correcting these errors to obtain an enhanced result. However, we observe that these studies are hardly comparable to each other, as they are usually evaluated on different corpora, paired with different ASR models, and even use different datasets to train the models. Accordingly, we first concentrate on providing an ASR hypothesis revising (HypR) dataset in this study. HypR contains several commonly used corpora (AISHELL-1, TED-LIUM 2, and LibriSpeech) and provides 50 recognition hypotheses for each speech utterance. The checkpoint models of ASR are also published. In addition, we implement and compare several classic and representative methods, showing the recent research progress in revising speech recognition results. We hope that the publicly available HypR dataset can become a reference benchmark for subsequent research and promote this field of research to an advanced level.