Amazon Web Services updated Amazon Lex’s speech recognition capabilities by launching a neural network-based automatic speech recognition (ASR) model for English. Trained on data from multiple English locales, the model demonstrates superior performance in recognizing conversational speech, including non-native accents and regional variations, reducing misrecognitions and user repetitions during voice interactions.
This enhancement aims to increase the effectiveness of self-service voice bots, improving customer experience in Amazon Connect and Lex applications. The feature can be enabled by selecting the 'Neural' recognition option in bot locale settings and is broadly available across AWS commercial regions supporting Lex and Connect services.
Source & further reading
Amazon Lex launches improved speech recognition models for English ↗Source published 2026-01-13. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.