Abstract
We describe KT-Speech-Crawler: an approach for automatic dataset construction for speech recognition by crawling YouTube videos. We outline several filtering and post-processing steps, which extract samples that can be used for training end-to-end neural speech recognition systems. In our experiments, we demonstrate that a single-core version of the crawler can obtain around 150 hours of transcribed speech within a day, containing an estimated 3.5% word error rate in the transcriptions. Automatically collected samples contain reading and spontaneous speech recorded in various conditions including background noise and music, distant microphone recordings, and a variety of accents and reverberation. When training a deep neural network on speech recognition, we observed around 40% word error rate reduction on the Wall Street Journal dataset by integrating 200 hours of the collected samples into the training set.- Anthology ID:
- D18-2016
- Volume:
- Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing: System Demonstrations
- Month:
- November
- Year:
- 2018
- Address:
- Brussels, Belgium
- Editors:
- Eduardo Blanco, Wei Lu
- Venue:
- EMNLP
- SIG:
- SIGDAT
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 90–95
- Language:
- URL:
- https://aclanthology.org/D18-2016
- DOI:
- 10.18653/v1/D18-2016
- Cite (ACL):
- Egor Lakomkin, Sven Magg, Cornelius Weber, and Stefan Wermter. 2018. KT-Speech-Crawler: Automatic Dataset Construction for Speech Recognition from YouTube Videos. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 90–95, Brussels, Belgium. Association for Computational Linguistics.
- Cite (Informal):
- KT-Speech-Crawler: Automatic Dataset Construction for Speech Recognition from YouTube Videos (Lakomkin et al., EMNLP 2018)
- PDF:
- https://preview.aclanthology.org/improve-issue-templates/D18-2016.pdf
- Code
- EgorLakomkin/KTSpeechCrawler