Adaptable Multi-Domain Language Model for Transformer ASR

Lee, Taewoo; Lee, Min-Joong; Kang, Tae Gyoon; Jung, Seokyeoung; Kwon, Minseok; Hong, Yeona; Lee, Jungin; Woo, Kyoung-Gu; Kim, Ho-Gyeong; Jeong, Jiseung; Lee, Ji‐Hyun; Lee, Hosik; Choi, Young Sang

doi:10.48550/arxiv.2008.06208

Cited by 2 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, one hot encoding of ground-truth domain information is also used when it is available during inference [2], [3], [7]. As for model based approaches, extra components, such as linear hidden network (LHN) [11], learning hidden unit contributions (LHUC) [12], [13], or adapters [14], are added to basic models to learn domain-specific information [7], [14]. Unfortunately, these approaches cannot be directly applied to our case since new songs may release at any time.…”

Section: Prior Workmentioning

confidence: 99%

A practical framework for multi-domain speech recognition and an instance sampling method to neural language modeling

Zhang¹,

Feng²,

Liu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Automatic speech recognition (ASR) systems used on smartphones or vehicles are usually required to process speechqueries from very different domains. In such situations, avanilla ASR system usually fails to perform well on every do-main. This paper proposes a multi-domain ASR frameworkfor Tencent Map, a navigation app used on smart phones andin-vehicle infotainment systems. The proposed frameworkconsists of three core parts: a basic ASR module to generaten-best lists of a speech query, a text classification moduleto determine which domain the speech query belongs to,and a reranking module to rescore n-best lists using domain-specific language models. In addition, an instance samplingbased method to training neural network language models(NNLMs) is proposed to address the data imbalance problemin multi-domain ASR. In experiments, the proposed frame-work was evaluated on navigation domain and music domain,since navigating and playing music are two main features ofTencent Map. Compared to a general ASR system, the pro-posed framework achieves a relative 13% ∼ 22% charactererror rate reduction on several test sets collected from TencentMap and our in-car voice assistant.

show abstract

Section: Prior Workmentioning

confidence: 99%

A practical framework for multi-domain speech recognition and an instance sampling method to neural language modeling

Zhang¹,

Feng²,

Liu³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Section: Prior Workmentioning

confidence: 99%

A practical framework for multi-domain speech recognition and an instance sampling method to neural language modeling

Zhang¹,

Feng²,

Liu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Automatic speech recognition (ASR) systems used on smart phones or vehiclesare usually required to process speech queries from very different domains.In such situations,a vanilla ASR system usually fails to perform well on every domain.This paper proposes a multi-domain ASR framework for Tencent Map,a navigation app used on smart phones and in-vehicle infotainment systems.The proposed framework consists of three core parts:a basic ASR module to generate n-best lists of a speech query,a text classification module to determine which domain the speech query belongs to,and a reranking module to rescore n-best lists using domain-specific language models.In addition,an instance sampling based method to training neural network language models (NNLMs)is proposed to address the data imbalance problem in multi-domain ASR.In experiments,the proposed framework was evaluated on navigation domain and music domain,since navigating and playing music are two main features of Tencent Map.Compared to a general ASR system,the proposed framework achieves a relative 13\% $\sim$ 22\% character error rate reductionon several test sets collected from Tencent Map and our in-car voice assistant.

show abstract

Adaptable Multi-Domain Language Model for Transformer ASR

Cited by 2 publications

References 0 publications

A practical framework for multi-domain speech recognition and an instance sampling method to neural language modeling

A practical framework for multi-domain speech recognition and an instance sampling method to neural language modeling

A practical framework for multi-domain speech recognition and an instance sampling method to neural language modeling

Contact Info

Product

Resources

About