mlsanityData to 7 Oct 2026

Text Injection for Neural Contextual Biasing

arXiv 2406.02921 ↗First posted 5 Jun 2024cs.CL cs.AI cs.LGcited by 3

Not on the map. Why?

Abstract (arXiv)

Neural contextual biasing effectively improves automatic speech recognition (ASR) for crucial phrases within a speaker's context, particularly those that are infrequent in the training data. This work proposes contextual text injection (CTI) to enhance contextual ASR. CTI leverages not only the paired speech-text data, but also a much larger corpus of unpaired text to optimize the ASR model and its biasing component. Unpaired text is converted into speech-like representations and used to guide the model's attention towards relevant bias phrases. Moreover, we introduce a contextual text-injected (CTI) minimum word error rate (MWER) training, which minimizes the expected WER caused by contextual biasing when unpaired text is injected into the model. Experiments show that CTI with 100 billion text sentences can achieve up to 43.3% relative WER reduction from a strong neural biasing model. CTI-MWER provides a further relative improvement of 23.5%.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 39 of its 47 references that are in the map. First posted on arXiv 5 Jun 2024.

Builds on

39 of 47 references.

References · 39

  1. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,853
  2. 2020Conformer: Convolution-augmented Transformer for Speech Recognitioncited by 1,838
  3. 2023Google USM: Scaling Automatic Speech Recognition Beyond 100 Languagescited by 228
  4. 2017State-of-the-art Speech Recognition With Sequence-to-Sequence Modelscited by 196
  5. 2020Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processingcited by 161
  6. 2018Deep context: end-to-end contextual speech recognitioncited by 136
  7. 2020Hybrid Autoregressive Transducer (hat)cited by 106
  8. 2020A Streaming On-Device End-to-End Model Surpassing Server-Side Conventional Model Quality and Latencycited by 99
  9. 2020Contextual RNN-T For Open Domain ASRcited by 85
  10. 2020Internal Language Model Estimation for Domain-Adaptive End-to-End Speech Recognitioncited by 80
  11. 2022MAESTRO: Matched Speech Text Representations through Modality Matchingcited by 79
  12. 2021SLAM: A Unified Encoder for Speech and Language Modeling via Speech-Text Joint Pre-Trainingcited by 69
27 more
  1. 2017Minimum Word Error Rate Training for Attention-based Sequence-to-Sequence Modelscited by 68
  2. 2022Unified Speech-Text Pre-training for Speech Translation and Recognitioncited by 64
  3. 2020Developing RNN-T Models Surpassing High-Performance Hybrid Models with Customization Capabilitycited by 63
  4. 2021Context-Aware Transformer Transducer for Speech Recognitioncited by 56
  5. 2021Internal Language Model Training for Domain-Adaptive End-to-End Speech Recognitioncited by 37
  6. 2021Tied & Reduced RNN-T Decodercited by 35
  7. 2020A New Training Pipeline for an Improved Neural Transducercited by 34
  8. 2021Factorized Neural Transducer for Efficient Language Model Adaptationcited by 31
  9. 2021Fast Contextual Adaptation with Neural Associative Memory for On-Device Personalized Speech Recognitioncited by 25
  10. 2022JOIST: A Joint Speech and Text Streaming Model For ASRcited by 24
  11. 2016Personalized Speech recognition on mobile devicescited by 23
  12. 2022Towards Reducing the Need for Speech Training Data To Build Spoken Language Understanding Systemscited by 20
  13. 2019RNN-T For Latency Controlled ASR With Improved Beam Searchcited by 19
  14. 2021Fast Text-Only Domain Adaptation of RNN-Transducer Prediction Networkcited by 19
  15. 2021Minimum Word Error Rate Training with Language Model Fusion for End-to-End Speech Recognitioncited by 19
  16. 2021Internal Language Model Adaptation with Text-Only Data for End-to-End Speech Recognitioncited by 17
  17. 2022Modular Hybrid Autoregressive Transducercited by 17
  18. 2020On Minimum Word Error Rate Training of the Hybrid Autoregressive Transducercited by 11
  19. 2023JEIT: Joint End-to-End Model and Internal Language Model Training for Speech Recognitioncited by 9
  20. 2023Contextual Biasing with the Knuth-Morris-Pratt Matching Algorithmcited by 9
  21. 2024Extreme Encoder Output Frame Rate Reduction: Improving Computational Latencies of Large End-to-End Modelscited by 5
  22. 2024Deferred NAM: Low-latency Top-K Context Injection via Deferred Context Encoding for Non-Streaming ASRcited by 5
  23. 2023Improving Joint Speech-Text Representations Without Alignmentcited by 4
  24. 2023Conversational Speech Recognition by Learning Audio-textual Cross-modal Contextual Representationcited by 4
  25. 2024Advanced Long-Content Speech Recognition With Factorized Neural Transducercited by 4
  26. 2023Massive End-to-end Models for Short Search Queriescited by 2
  27. 2023Text Injection for Capitalization and Turn-Taking Prediction in Speech Modelscited by 1

Built on by

Cited by year

Citing papers per year: 2024: 1; 2025: 2.

3 papers.

  1. 2024Optimizing Contextual Speech Recognition Using Vector Quantization for Efficient Retrievalcited by 3
  2. 2025Retrieval Augmented Generation based context discovery for ASRcited by 1
  3. 2025A Neural Model for Contextual Biasing Score Learning and Filteringcited by 1