Google’s Universal Speech Model a “critical first step” towards 1,000 languages
Google has taken its first step towards bringing its 1,000 Languages Initiative to fruition, with the company sharing details on its Universal Speech Model (USM).
Google describes the USM as “a critical first step” towards supporting 1,000 languages. The company announced the 1,000 Languages Initiative in November 2022.
Through the initiative, Google aims to build a machine learning model to support 1,000 of the world’s most-spoken languages.
The tech giant says USM is a family of speech models with two billion parameters trained on 12 million hours of speech and 28 billion sentences of text.
It spans over 300 languages and is currently used for closed YouTube captions.
“[USM] can perform automatic speech recognition (ASR) not only on widely-spoken languages like English and Mandarin, but also on under-resourced languages like Amharic, Cebuano, Assamese, and Azerbaijani to name a few,” says Google.
The tech giant says USM currently supports more than 100 languages and will serve as the foundation to build on which to build a more comprehensive system.
“We believe USM’s base model architecture and training pipeline comprise a foundation on which we can build to expand speech modelling to the next 1,000 languages,” the company says.
It is not entirely clear how Google plans to use the technology. However, one use could be in augmented reality glasses for real-time translations.
Now read: OpenAI now selling ChatGPT integration to businesses
Loading ...