要旨

本稿では，ユーザが任意に登録したフレーズに対する認識精度を改善するため，編集可能な用語リストを用いた音声認識モデルのカスタマイズ手法を提案する．提案手法では，従来のConnectionist Temporal Classification/attention型音声認識モデルにテキストエンコーダとCross-attention層を追加し，音声信号に含まれる登録用語を検出する．また，登録用語の認識精度向上のため，検出された登録単語の確率やConnectionist Temporal Classificationを統合した推論アルゴリズムを提案する．日本語の社内データセットおよび英語のLibrispeechコーパスを用いた評価の結果，提案手法の適用により，登録用語の文字誤り率と単語誤り率が改善した．

参考文献

1. （１） Prabhavalkar, R., Hori, T., Sainath, T. N., Schlüter, R., Watanabe, S.: End-to-End Speech Recognition: A Survey, IEEE/ACM Transactions on Audio, Speech, and Language Processing, Vol. 32, p. 325-351, (2023), doi: 10.1109/TASLP.2023.3328283
2. （２） https://global.honda/jp/stories/003.html, （参照2024/9/23）
3. （３） https://www.jp.honda-ri.com/activity/honda-ca-system/, （参照2024/9/23）
4. （４） Graves, A., Fernández, S., Gomez, F., Schmidhuber, J.: Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks, ICML ‘06: Proceedings of the 23rd international conference on Machine learning, p. 369-376, (2006), doi: 10.1145/1143844.1143891
5. （５） Peng, Y., Sudo, Y., Shakeel, M., Watanabe, S.: OWSM-CTC: An Open Encoder-Only Speech Foundation Model for Speech Recognition, Translation, and Language Identification, Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics, Vol. 1, p. 10192–10209, (2024), doi: 10.18653/v1/2024.acl-long.549
6. （６） Graves, A.: Sequence Transduction with Recurrent Neural Networks, International Conference of Machine Learning (ICML) 2012 Workshop on Representation Learning, arXiv preprint, (2012), doi: 10.48550/arXiv.1211.3711
7. （７） Gulati, A., Qin, J., Chiu, C.-C., Parmar, N., Zhang, Y., Yu, J., Han, W., Wang, S., Zhang, Z., Wu, Y., Pang, R.: Conformer: Convolution-augmented Transformer for Speech Recognition, Proc. Interspeech 2020, p. 5036-5040, (2020), doi: 10.21437/Interspeech.2020-3015
8. （８） Chorowski, J. K., Bahdanau, D., Serdyuk, D., Cho, K., Bengio, Y.: Attention-Based Models for Speech Recognition, Advances in Neural Information Processing Systems, Vol. 28, (2015)
9. （９） Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., Sutskever, I.: Robust Speech Recognition via Large-Scale Weak Supervision, Proceedings of the 40th International Conference on Machine Learning, p. 28492-28518, (2023)
10. （10） Watanabe, S., Hori, T., Kim, S., Hershey, J. R., Hayashi, T.: Hybrid CTC/Attention Architecture for End-to-End Speech Recognition, IEEE Journal of Selected Topics in Signal Processing, Vol. 11, Issue 8, p. 1240-1253, (2017), doi: 10.1109/JSTSP.2017.2763455
11. （11） Sudo, Y., Shakeel, M., Yan, B., Shi, J., Watanabe, S.: 4D ASR: Joint modeling of CTC, Attention, Transducer, and Mask-Predict decoders, Proc. INTERSPEECH 2023, p. 3312-3316, (2023), doi: 10.21437/Interspeech.2023-1303
12. （12） Pundak, G., Sainath, T. N., Prabhavalkar, R., Kannan, A., Zhao, D.: Deep Context: End-to-end Contextual Speech Recognition, 2018 IEEE Spoken Language Technology Workshop (SLT), (2018), doi: 10.1109/SLT.2018.8639034
13. （13） Huang, K., Zhang, A., Yang, Z., Guo, P., Mu, B., Xu, T., Xie, L.: Contextualized End-to-End Speech Recognition with Contextual Phrase Prediction Network, Proc. INTERSPEECH 2023, p. 4933-4937, (2023), doi: 10.21437/Interspeech.2023-767
14. （14） Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., Polosukhin, I.: Attention is all you need, NIPS’17: Proceedings of the 31st International Conference on Neural Information Processing Systems, p. 6000-6010, (2017)
15. （15） Shakeel, M., Sudo, Y., Peng, Y., Watanabe, S.: Contextualized End-to-end Automatic Speech Recognition with Intermediate Biasing Loss, Proc. Interspeech 2024, p. 3909-3913, (2024), doi: 10.21437/Interspeech.2024-1257
16. （16） Huber, C., Hussain, J., Stüker, S., Waibel, A.: Instant One-Shot Word-Learning for Context-Specific Neural Sequence-to-Sequence Speech Recognition, 2021 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), (2021), doi: 10.1109/ASRU51503.2021.9687898
17. （17） Sudo, Y., Shakeel, M., Fukumoto, Y., Peng, Y., Watanabe, S.: Contextualized Automatic Speech Recognition With Attention-Based Bias Phrase Boosted Beam Search, ICASSP 2024 – 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), p. 10896-10900, (2024), doi: 10.1109/ICASSP48485.2024.10447782
18. （18） Panayotov, V., Chen, G., Povey, D., Khudanpur, S.: Librispeech: An ASR corpus based on public domain audio books, 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), p. 5206-5210, (2015), doi: 10.1109/ICASSP.2015.7178964
19. （19） Maekawa, K.: Corpus of Spontaneous Japanese: Its design and evaluation, Proc. ISCA/IEEE Workshop on Spontaneous Speech Processing and Recognition, (2003), https://www.isca-archive.org/sspr_2003/maekawa03_sspr.pdf, （参照2024/9/23）
20. （20） Kurematsu, A., Takeda, K., Sagisaka, Y., Katagiri, S., Kuwabara, H., Shikano, K.: ATR Japanese speech database as a tool of speech recognition and synthesis, Speech Communication, Vol. 9, Issue 4, p. 357-363, (1990), doi: 10.1016/0167-6393(90)90011-W
21. （21） Sudo, Y., Hata, K., Nakadai, K.: Retraining-free Customized ASR for Enharmonic Words Based on a Named-Entity-Aware Model and Phoneme Similarity Estimation, Proc. INTERSPEECH 2023, p. 491-495, (2023), doi: 10.21437/Interspeech.2023-1320

著者(所属)

• 周藤 唯（株式会社ホンダ・リサーチ・インスティチュート・ジャパン）
• Shakeel MUHAMMAD（株式会社ホンダ・リサーチ・インスティチュート・ジャパン）
• 住田 直亮（株式会社ホンダ・リサーチ・インスティチュート・ジャパン）

掲載情報

Honda R&D Technical Review　Vol.38 収録 (135 - 142)

論文PDF

引用情報