/KTSpeechCrawler

Primary LanguagePythonMIT LicenseMIT

(to be updated with-in next couple of days)

KT-Speech-Crawler: Automatic Dataset Construction for Speech Recognition from YouTube Videos

Web demo

http://emnlp-demo.lakomkin.me/

Installation

git clone https://github.com/EgorLakomkin/KTSpeechCrawler
pip install webvtt-py SpeechRecognition path.py flask python-Levenshtein tqdm unicodedata

Docker

TODO:

Running crawler

sh ./crawler/en_corpus.sh <dir_with_intermediate_results> <dir_for_resulting_samples>

Browsing samples

python server.py --corpus <dir_for_resulting_samples>
Goto: http://localhost:8888/

Citation

@article{lakomkin2018kt, title={KT-Speech-Crawler: Automatic Dataset Construction for Speech Recognition from YouTube Videos}, author={Lakomkin, Egor and Magg, Sven and Weber, Cornelius and Wermter, Stefan}, journal={EMNLP 2018}, pages={90}, year={2018} }