Facebook AI software does speech recognition without any transcribed data
Facebook AI software does speech recognition without any transcribed data
High-performance speech recognition with no supervision at all
Facebook AI post https://ai.facebook.com/blog/wav2vec-unsupervised-speech-recognition-without-supervision
Demo in Twitter video https://twitter.com/schrep/status/1395766932104572928
Unsupervised Speech Recognition
Facebook Paper abstract https://ai.facebook.com/research/publications/unsupervised-speech-recognition
Facebook PDF paper https://scontent.ftpa1-2.fna.fbcdn.net/v/t39.8562-6/187874612_311717527241594_5668815448923437055_n.pdf?_nc_cat=102&ccb=1-3&_nc_sid=ae5e01&_nc_ohc=qXOFpESwsyUAX_klfAQ&_nc_ht=scontent.ftpa1-2.fna&oh=b5e204449e96aed9ad804fde85d3ad72&oe=60D1D515
GitHub https://github.com/pytorch/fairseq/tree/master/examples/wav2vec/unsupervised
… current speech recognition technology still requires labeled training data. This limits this technology to a small fraction of the languages spoken around the globe.
This paper describes … method to train speech recognition models without any labeled data.
… wav2vec-U reduces the phoneme error rate on the TIMIT benchmark from 26.1 to 11.3.
… wav2vec-U … rivaling some of the best published systems trained on 960 hours of labeled data from only two years ago.
We also experiment on nine other languages, including low-resource languages such as Kyrgyz, Swahili and Tatar.
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
