A Likelihood Ratio Based Domain Adaptation Method for E2E Models

Choudhury, Chhavi; Gandhe, Ankur; Ding, Xiaoyan; Bulyko, Ivan

doi:10.1109/icassp43922.2022.9746480

Cited by 8 publications

(1 citation statement)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to a large amount of labelled training data, E2E ASR models surpass pipeline methods on most public datasets [3]. However, E2E ASR still suffers from unseen domains [4], and large quantities of labelled data are not always feasible to collect and can therefore be limited [5]. Adaptation training methods can be utilised to alleviate this issue when the target domain has enough paired data [6,7].…”

Section: Introductionmentioning

confidence: 99%

Adaptable End-to-End ASR Models using Replaceable Internal LMs and Residual Softmax

Deng¹,

Woodland²

2023

Preprint

View full text Add to dashboard Cite

End-to-end (E2E) automatic speech recognition (ASR) implicitly learns the token sequence distribution of paired audiotranscript training data. However, it still suffers from domain shifts from training to testing, and domain adaptation is still challenging. To alleviate this problem, this paper designs a replaceable internal language model (RILM) method, which makes it feasible to directly replace the internal language model (LM) of E2E ASR models with a target-domain LM in the decoding stage when a domain shift is encountered. Furthermore, this paper proposes a residual softmax (R-softmax) that is designed for CTC-based E2E ASR models to adapt to the target domain without re-training during inference. For E2E ASR models trained on the LibriSpeech corpus, experiments showed that the proposed methods gave a 2.6% absolute WER reduction on the Switchboard data and a 1.0% WER reduction on the AESRC2020 corpus while maintaining intra-domain ASR results.

show abstract