AWS’ Amazon Transcribe Now Supports 100 Languages with Generative AI

AWS has introduced new enhancements to its Amazon Transcribe service, equipping it with generative AI-based transcription capabilities for 100 languages. This update includes a range of new AI features tailored to meet the diverse needs of customers.

During the AWS re: Invent event, it was announced that Amazon Transcribe has expanded its language recognition capabilities and can now transcribe calls in addition to its existing functions. AWS customers leverage Transcribe to integrate speech-to-text capabilities into their applications on the AWS Cloud.

In a blog post, the company explained that Transcribe underwent training on an extensive volume of unlabeled audio data across 100 languages. This enabled the system to utilize self-supervised algorithms to recognize speech patterns across various languages and accents. AWS ensured that less common languages received equal representation in the training data, emphasizing accuracy across all languages.

At the end of 2022, Amazon Transcribe had support for 79 languages, but with this update, it now covers 100 languages. AWS states that Amazon Transcribe achieves an accuracy rate of 20-50% across multiple languages. Additionally, it provides automatic punctuation, custom vocabulary, automatic language identification, and custom vocabulary filters. The tool is designed to recognize speech in both audio and video formats, even in noisy environments.

The Verge reached out to AWS to gather information on the previous accuracy of Amazon Transcribe and the foundation models used by the platform.

Furthermore, AWS emphasized that the improvements in language recognition also enhance the accuracy of its Call Analytics platform, benefiting its contact center customers. With generative AI models, Amazon Transcribe Call Analytics can now provide summaries of interactions between agents and customers. This functionality reduces post-call work for creating reports, enabling managers to quickly access relevant information without the need to review the entire transcript.

While AWS is a prominent provider of AI-powered transcription services, it faces competition from other companies such as Otter, which offers AI transcriptions and released a summarization tool in June. Meta is also working on a generative AI-powered translation model capable of recognizing nearly 100 spoken languages, setting the stage for increased competition in this space.

Additionally, AWS has introduced new capabilities to its Amazon Personalization product, allowing clients to offer tailored product recommendations to customers, akin to how streaming services suggest new content based on viewing history. The addition of Content Generation enables the platform to create thematically linked titles or email subject lines for recommendation lists.

Source link

Leave a Comment