Amazon recently expanded the language options in its Amazon Transcribe service. Now, it can transcribe spoken words in 100 different languages using advanced AI technology. This was revealed at the AWS re:Invent event.
The updated Amazon Transcribe also comes with additional AI features. It allows AWS customers to turn spoken words into text in various languages and facilitates call transcription. AWS customers typically use Transcribe to incorporate speech-to-text functions into their applications on the AWS Cloud.
Improvements in Amazon Transcribe
According to a blog post by the company, Transcribe was trained on a vast amount of unlabeled audio data from over 100 languages, totaling millions of hours. It employs self-supervised algorithms to understand patterns in human speech across various languages and accents. AWS took care to avoid over-representing certain languages in the training data, ensuring that less commonly spoken languages could be transcribed as accurately as more frequently spoken ones.
By the end of 2022, Amazon Transcribe could handle 79 languages.
AWS reports that Amazon Transcribe achieves an accuracy rate ranging from 20 percent to 50 percent across numerous languages. It provides features such as automatic punctuation, custom vocabulary, automatic language identification, and custom vocabulary filters. This tool can accurately transcribe speech in both audio and video formats, even in noisy environments.
Improvements in language recognition not only enhance Amazon Transcribe but also contribute to increased accuracy in the Call Analytics platform. This platform is frequently used by contact center customers.
The newly enhanced Amazon Transcribe Call Analytics, now powered by generative AI models, summarizes interactions between agents and customers. This advancement reduces the post-call workload of creating reports, allowing managers to quickly access information without having to review the entire transcript.
AI-Powered Transcription Services
Certainly, AWS isn’t the sole provider of AI-driven transcription services. Otter has been delivering AI transcriptions to both consumers and enterprises for some time and introduced a summarization tool in June. Although not identical, Meta revealed ongoing efforts to develop a generative AI-powered translation model that can understand nearly 100 spoken languages.
Furthermore, AWS expanded the capabilities of its Amazon Personalization product. This tool enables clients to present products or recommend items to customers, similar to how streaming services suggest new shows based on past user activity. AWS introduced Content Generation as part of this, a feature that generates titles or email subject lines to thematically align with recommendation lists. Amazon is now providing free educational courses on generative AI.
Conclusion
Amazon Transcribe’s language expansion and AI enhancements mark a significant stride in speech-to-text technology. With improvements in accuracy and new features, including advanced call analytics, it solidifies its position in a competitive field. AWS’s commitment to innovation extends to other services like personalization, showcasing the broader impact of AI in diverse applications.
