[ad_1]
Amazon has announced that its AI-powered Transcribe service is now available with better accuracy support for over 100 languages, including some Indian languages. The company said that its ‘next-generation multi-billion parameter speech foundation model-powered system’ helped in the expansion of its automatic speech recognition support.
Amazon Transcribe is an automatic speech recognition (ASR) service that makes it easy for users to add speech-to-text capabilities apps.
The company said that the service’s speech foundation model is trained using self-supervised algorithms that help it learn inherent universal patterns of human speech across languages and accents.
“It is trained on millions of hours of unlabeled audio data from over 100 languages. The training recipes are optimised through smart data sampling to balance the training data between languages, ensuring that traditionally under-represented languages also reach high accuracy levels,” Amazon said.
Indian languages support
Among the over 100 languages, Amazon Transcribe supports Bengali, English (Indian), Gujarati, Hindi, Kannada, Malayalam, Marathi, Punjabi, Tamil and Telugu Indian languages.
The e-commerce giant said that by leveraging the speech foundation model, Amazon Transcribe is able to deliver ‘significant accuracy improvement between 20% and 50% across most languages.’
On telephony speech, accuracy improvement is between 30% and 70%, Amazon said, adding that the ASR model also delivers improvements in readability with more accurate punctuation and capitalisation.
Meta’s AI translation model
Earlier this year, Meta also announced a generative AI-powered translation model that recognises nearly 100 spoken languages. Called SeamlessM4T, which stands for Massively Multilingual and Multimodal Machine Translation, the model is said to have the capability to translate speech-to-text and text-to-text for nearly 100 languages into 35 output languages.
Meanwhile, Google’s Speech-to-Text language support is available for over 125 languages and variants.
Amazon Transcribe is an automatic speech recognition (ASR) service that makes it easy for users to add speech-to-text capabilities apps.
The company said that the service’s speech foundation model is trained using self-supervised algorithms that help it learn inherent universal patterns of human speech across languages and accents.
“It is trained on millions of hours of unlabeled audio data from over 100 languages. The training recipes are optimised through smart data sampling to balance the training data between languages, ensuring that traditionally under-represented languages also reach high accuracy levels,” Amazon said.
Indian languages support
Among the over 100 languages, Amazon Transcribe supports Bengali, English (Indian), Gujarati, Hindi, Kannada, Malayalam, Marathi, Punjabi, Tamil and Telugu Indian languages.
The e-commerce giant said that by leveraging the speech foundation model, Amazon Transcribe is able to deliver ‘significant accuracy improvement between 20% and 50% across most languages.’
On telephony speech, accuracy improvement is between 30% and 70%, Amazon said, adding that the ASR model also delivers improvements in readability with more accurate punctuation and capitalisation.
Meta’s AI translation model
Earlier this year, Meta also announced a generative AI-powered translation model that recognises nearly 100 spoken languages. Called SeamlessM4T, which stands for Massively Multilingual and Multimodal Machine Translation, the model is said to have the capability to translate speech-to-text and text-to-text for nearly 100 languages into 35 output languages.
Meanwhile, Google’s Speech-to-Text language support is available for over 125 languages and variants.
[ad_2]
Source link
More Stories
Google Maps: Three privacy features coming to Google Maps on Android, iPhones
Most-Downloaded IPhone App: This Chinese app was the most-downloaded iPhone app in the US in 2023
Ukraine’s largest mobile operator goes offline for millions of users after cyber attack