Acoustic Data-Driven Subword Modeling for End-to-End Speech Recognition

Zhou, Wei; Zeineldeen, Mohammad; Zheng, Zuoyun; Schlüter, Ralf; Ney, Hermann

doi:10.21437/interspeech.2021-1623

Cited by 9 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Whole word models [30], according to Zipf's law [31], would require unrealistically high amounts of transcribed training data for large vocabularies, which might not be attainable for many tasks. On the other hand, methods to generate subword vocabularies based on characters, like the currently popular byte pair encoding (BPE) approach [32], might be seen as secondary approaches outside the E2E objective, even more so if acoustic data is considered for subword derivation [33], [34], [35], [36].…”

Section: Distinctiveness Of the Term E2ementioning

confidence: 99%

“…via byte-pair encoding [32]. Also, available pronunciation lexica can be utilized indirectly for assisting subword generation for E2E systems [35], [36], which are shown to outperform byte-pair encoding. Within classical ASR systems, phonetic clustering also can be avoided completely by modeling phonemes in context directly [220].…”

Section: Relationship To Classical Asrmentioning

confidence: 99%

See 1 more Smart Citation

End-to-End Speech Recognition: A Survey

Prabhavalkar,

Hori,

Sainath

et al. 2024

IEEE/ACM Trans. Audio Speech Lang. Process.

Self Cite

View full text Add to dashboard Cite

In the last decade of automatic speech recognition (ASR) research, the introduction of deep learning has brought considerable reductions in word error rate of more than 50% relative, compared to modeling without deep learning. In the wake of this transition, a number of all-neural ASR architectures have been introduced. These so-called end-to-end (E2E) models provide highly integrated, completely neural ASR models, which rely strongly on general machine learning knowledge, learn more consistently from data, with lower dependence on ASR domainspecific experience. The success and enthusiastic adoption of deep learning, accompanied by more generic model architectures has led to E2E models now becoming the prominent ASR approach. The goal of this survey is to provide a taxonomy of E2E ASR models and corresponding improvements, and to discuss their properties and their relationship to classical hidden Markov model (HMM) based ASR architectures. All relevant aspects of E2E ASR are covered in this work: modeling, training, decoding, and external language model integration, discussions of performance and deployment opportunities, as well as an outlook into potential future developments.

show abstract

Section: Distinctiveness Of the Term E2ementioning

confidence: 99%

Section: Relationship To Classical Asrmentioning

confidence: 99%

End-to-End Speech Recognition: A Survey

Prabhavalkar,

Hori,

Sainath

et al. 2024

IEEE/ACM Trans. Audio Speech Lang. Process.

Self Cite

View full text Add to dashboard Cite

show abstract

“…via byte-pair encoding [27]. Also, available pronunciation lexica can be utilized indirectly for assisting subword generation for E2E systems [290], [291], which are shown to outperform byte-pair encoding. Within classical ASR systems, phonetic clustering also can be avoided completely by modeling phonemes in context directly [292].…”

Section: Use Of Large-scale Pretrained Lmsmentioning

confidence: 99%

End-to-End Speech Recognition: A Survey

Prabhavalkar¹,

Hori²,

Sainath³

et al. 2023

Preprint

View full text Add to dashboard Cite

In the last decade of automatic speech recognition (ASR) research, the introduction of deep learning brought considerable reductions in word error rate of more than 50% relative, compared to modeling without deep learning. In the wake of this transition, a number of all-neural ASR architectures were introduced. These so-called end-to-end (E2E) models provide highly integrated, completely neural ASR models, which rely strongly on general machine learning knowledge, learn more consistently from data, while depending less on ASR domainspecific experience. The success and enthusiastic adoption of deep learning accompanied by more generic model architectures lead to E2E models now becoming the prominent ASR approach. The goal of this survey is to provide a taxonomy of E2E ASR models and corresponding improvements, and to discuss their properties and their relation to the classical hidden Markov model (HMM) based ASR architecture. All relevant aspects of E2E ASR are covered in this work: modeling, training, decoding, and external language model integration, accompanied by discussions of performance and deployment opportunities, as well as an outlook into potential future developments.

show abstract

“…[8] and the 300h SWB corpus [9]. We evaluate the proposed training pipeline on context-1 transducer models using phonemes for both corpora, and full-context transducer models using 5k acoustic data-driven subword modeling (ADSM) units [34] for LBS. Additionally, we reduce the LBS phoneme inventory in the official lexicon by unifying stressed phonemes, e.g.…”

Section: Learning Rate Scheduling and Epochsmentioning

confidence: 99%

“…By default, we apply 1-pass LM SF decoding, where word-level LM is used for phoneme transducers. The word-level transformer (Trafo) LMs are the same as in [38] for LBS and [39] (sentence-wise) for SWB, while the ADSM Trafo LM is the same as in [34].…”

Section: Learning Rate Scheduling and Epochsmentioning

confidence: 99%

Efficient Training of Neural Transducer for Speech Recognition

Zhou¹,

Michel²,

Schlüter³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

As one of the most popular sequence-to-sequence modeling approaches for speech recognition, the RNN-Transducer has achieved evolving performance with more and more sophisticated neural network models of growing size and increasing training epochs. While strong computation resources seem to be the prerequisite of training superior models, we try to overcome it by carefully designing a more efficient training pipeline. In this work, we propose an efficient 3-stage progressive training pipeline to build highly-performing neural transducer models from scratch with very limited computation resources in a reasonable short time period. The effectiveness of each stage is experimentally verified on both Librispeech and Switchboard corpora. The proposed pipeline is able to train transducer models approaching state-of-the-art performance with a single GPU in just 2-3 weeks. Our best conformer transducer achieves 4.1% WER on Librispeech test-other with only 35 epochs of training.

show abstract

Acoustic Data-Driven Subword Modeling for End-to-End Speech Recognition

Cited by 9 publications

References 0 publications

End-to-End Speech Recognition: A Survey

End-to-End Speech Recognition: A Survey

End-to-End Speech Recognition: A Survey

Efficient Training of Neural Transducer for Speech Recognition

Contact Info

Product

Resources

About