Facebook Parent Company Meta Uses Bible Translations To Train AI Speech Recognition
Unsplash image by Jenny Smith
The artificial intelligence arm of Facebook parent company Meta has turned to widely translated religious texts like the Bible to provide data for its Massively Multilingual Speech project, which aims to expand text-to-speech and speech-to-text technology to underserved languages.
The company said that equipping machines with the ability to recognize and produce speech can make information accessible to more people, including those who rely entirely on information that’s delivered verbally. Existing speech recognition models only cover approximately 100 languages, less than 2% of the more than 7,000 known languages.
Meta tapped text-based Bible translation research and audio recordings of people reading the Bible to create a dataset of readings of the New Testament in more than 1,100 languages, which it said provided an average of 32 hours of data per language.
The MMS project also produced lesser data for nearly 4,000 languages, some of which have only a few hundred speakers.
Meta said that while 32 hours of data per language is not enough to train conventional supervised speech recognition models, it is building new models on its self-supervised speech representation learning platform.
Researchers said that although the content of the audio recordings used for the “training” was religious, it did not overly bias the technology to produce more religious language.
Meta is publicly sharing its models and code to allow others in the research community to build on its work.
“Many of the world’s languages are in danger of disappearing, and the limitations of current speech recognition and speech generation technology will only accelerate this trend,” the company said. “We envision a world where technology has the opposite effect, encouraging people to keep their languages alive since they can access information and use technology by speaking in their preferred language.”
Anne Stych is a freelance writer, copy editor, proofreader and content manager covering science, technology, retail, etc. She writes for American City Business Journals’ BizWomen. This story originally appeared at MinistryWatch.