Posted On: Mar 11, 2021

Amazon Transcribe is an automatic speech recognition (ASR) service that makes it easy for developers to add speech-to-text capabilities to their voice-enabled applications. Up until now, Amazon Transcribe has offered confidence scores for each word in its transcription output for its batch (asynchronous) API. Starting today, Amazon Transcribe will surface word-level confidence scores for its streaming (synchronous) transcription API as well.

Confidence scores are the service’s self-evaluations on how well it may have transcribed a word and should not be confused with accuracy measurements. Each word has an associated confidence score between 0 and 1. These scores offer developers a range of opportunities to implement post-processing logic on text transcripts. For instance, you may only want to use terms that meet a specified confidence threshold for any post-processing like keyword search, translation or human review. Whatever your voice application may be, confidence scores provide you with an additional tool to augment or consume transcripts for your particular streaming audio use case. Common use cases for streaming transcriptions includes real-time call transcriptions and live captions for broadcast media.
Word-level confidence score for streaming transcription is available in all AWS Regions where Amazon Transcribe streaming API is available. Visit our service console today to try it out. You can learn more by checking out the technical documentation.