mlsanityData to 14 Oct 2026

On lattice-free boosted MMI training of HMM and CTC-based full-context ASR models

arXiv 2107.04154 ↗First posted 9 Jul 2021eess.AS cs.LGcited by 7

Not on the map. Why?

Abstract (arXiv)

Hybrid automatic speech recognition (ASR) models are typically sequentially trained with CTC or LF-MMI criteria. However, they have vastly different legacies and are usually implemented in different frameworks. In this paper, by decoupling the concepts of modeling units and label topologies and building proper numerator/denominator graphs accordingly, we establish a generalized framework for hybrid acoustic modeling (AM). In this framework, we show that LF-MMI is a powerful training criterion applicable to both limited-context and full-context models, for wordpiece/mono-char/bi-char/chenone units, with both HMM/CTC topologies. From this framework, we propose three novel training schemes: chenone(ch)/wordpiece(wp)-CTC-bMMI, and wordpiece(wp)-HMM-bMMI with different advantages in training performance, decoding efficiency and decoding time-stamp accuracy. The advantages of different training schemes are evaluated comprehensively on Librispeech, and wp-CTC-bMMI and ch-CTC-bMMI are evaluated on two real world ASR tasks to show their effectiveness. Besides, we also show bi-char(bc) HMM-MMI models can serve as better alignment models than traditional non-neural GMM-HMMs.

Key facts. Cited by 7 papers in the map (data through 14 Oct 2026). Builds on 20 of its 43 references that are in the map. First posted on arXiv 9 Jul 2021.

Builds on

20 of 43 references.

References · 20

  1. 2018SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processingcited by 1,989
  2. 2020Conformer: Convolution-augmented Transformer for Speech Recognitioncited by 1,863
  3. 2019SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognitioncited by 1,380
  4. 2013Speech Recognition with Deep Recurrent Neural Networkscited by 669
  5. 2015Listen, Attend and Spellcited by 543
  6. 2017State-of-the-art Speech Recognition With Sequence-to-Sequence Modelscited by 196
  7. 2015End-to-End Attention-based Large Vocabulary Speech Recognitioncited by 151
  8. 2020Emformer: Efficient Memory Transformer Based Acoustic Model For Low Latency Streaming Speech Recognitioncited by 113
  9. 2021Contextualized Streaming End-to-End Speech Recognition with Trie-Based Deep Biasing and Shallow Fusioncited by 101
  10. 2020Alignment Restricted Streaming Recurrent Neural Network Transducercited by 42
  11. 2020A New Training Pipeline for an Improved Neural Transducercited by 34
  12. 2021Why does CTC result in peaky behavior?cited by 33
8 more
  1. 2020Improving RNN Transducer Based ASR with Auxiliary Taskscited by 27
  2. 2019From Senones to Chenones: Tied Context-Dependent Graphemes for Hybrid Speech Recognitioncited by 20
  3. 2020PyChain: A Fully Parallelized PyTorch Implementation of LF-MMI for End-to-End ASRcited by 16
  4. 2020Benchmarking LF-MMI, CTC and RNN-T Criteria for Streaming ASRcited by 11
  5. 2020Faster, Simpler and More Accurate Hybrid ASR Systems Using Wordpiecescited by 9
  6. 2018Promising Accurate Prefix Boosting for sequence-to-sequence ASRcited by 4
  7. 2018Sequence Discriminative Training for Deep Learning based Acoustic Keyword Spottingcited by 1
  8. 2018A Comparison of Lattice-free Discriminative Training Criteria for Purely Sequence-Trained Neural Network Acoustic Modelscited by 1

Built on by

Cited by year

Citing papers per year: 2022: 3; 2023: 2; 2024: 2.

7 papers.

  1. 2022Supervision-Guided Codebooks for Masked Prediction in Speech Pre-trainingcited by 14
  2. 2024Less Peaky and More Accurate CTC Forced Alignment by Label Priorscited by 12
  3. 2022Integrating Lattice-Free MMI into End-to-End Speech Recognitioncited by 3
  4. 2022Lattice-Free Sequence Discriminative Training for Phoneme-Based Neural Transducerscited by 3
  5. 2024The Conformer Encoder May Reverse the Time Dimensioncited by 2
  6. 2023End-to-End Training of a Neural HMM with Label and Transition Probabilitiescited by 1
  7. 2023Investigating the Effect of Language Models in Sequence Discriminative Training for Neural Transducerscited by 0