Unveiled during the re:Invent 2023 conference, the upgraded features encompass Amazon Transcribe’s introduction of foundational model-powered language support and AI-enhanced call analytics. Amazon Personalize leverages foundation model for the creation of more comprehensive content recommendations, while Amazon Lex utilizes large language models to deliver precise and engaging conversational responses.
The revamped Amazon Transcribe, now enhanced with foundation model technology, has a noteworthy increase in accuracy, ranging from 20% to 50% across the majority of languages, according to AWS. This upgraded automatic speech recognition (ASR) system introduces distinctive features for over 100 supported languages, focusing on aspects such as user-friendly interface, customization options, and enhanced user safety and privacy.
Some of the highlighted features are automatic punctuation, personalized vocabulary options, automatic language identification, speaker diarization (the ability to identify and distinguish different speakers in an audio recording), word-level confidence scores, and custom vocabulary filters. With support for a diverse array of languages and these value-added features, enterprises are positioned to extract meaningful insights from their audio content.
Amazon Personalize, the machine learning service crafted to assist developers in creating personalized recommendations, has elevated its capabilities with hyper-personalization through a feature named Content Generator, powered by foundational models. Utilizing natural language, the latest feature generates straightforward and compelling text that outlines the thematic links among recommended items. As per AWS, this functionality empowers companies to automatically produce captivating titles or email subject lines, enticing customers to click on videos or make purchases.
AWS is making Personalize compatible with the open-source LangChain framework, enabling customers to construct their own foundational model-based applications. This integration allows users to seamlessly incorporate Amazon Personalize, obtain recommendations for a campaign or recommender, and seamlessly integrate them into their foundational model-powered applications within the LangChain ecosystem.
Amazon Lex, the fully managed AI service for integrating conversational interfaces into applications using voice and text, is now equipped with foundational model-powered capabilities. This enhancement facilitates quicker bot development and enhances containment within Amazon’s AI service. Amazon Lex introduces Conversation FAQ (CFAQ), a novel capability designed to intelligently address frequently asked customer questions on a large scale. Fueled by foundational models sourced from Amazon Bedrock and validated knowledge bases, CFAQ empowers companies to deliver precise and automated responses to typical customer queries in a natural and engaging manner.
