AWS's transcription platform can now recognize 100 spoken languages
Amazon Web Services (AWS) has revealed that its transcription platform, Amazon Transcribe, is now powered by generative AI models. The update, announced at the AWS re:Invent event, states that Transcribe has been trained on "millions of hours of unlabeled audio data from over 100 languages" using self-supervised algorithms. Till late 2022, Transcribe supported only 79 languages. However, the platform can now recognize 100 spoken languages and spin up a call transcription. It also gets a host of new AI capabilities.
Transcribe's accuracy improved by up to 50%
Thanks to the latest advancements, Amazon Transcribe now has 20% to 50% better accuracy across many languages, according to AWS. The platform also provides automatic punctuation, automatic language identification, and custom vocabulary filters. It can even recognize speech in audio and video formats, as well as in noisy environments. These improvements also boost the accuracy of AWS's Call Analytics platform, which is often used by contact center customers to summarize interactions between agents and clients.
Competitors and additional capabilities
AWS faces competition in the AI-powered transcription market from companies like Otter, which has been offering similar services to consumers and enterprises for a while now. Meta is also developing a generative AI-powered translation model that recognizes nearly 100 spoken languages. In addition to the transcription enhancements, AWS announced new capabilities for its Amazon Personalization product, including Content Generation, which writes titles or email subject lines to thematically connect recommendation lists for clients.
