A Script and Tutorial for Using Rev AI's Automatic Speech Transcription

Abstract

We introduce Speech Transcriber with Rev AI (STR) - a Python script that allows for easy interfacing with the Rev AI speech transcription service. Recent advancements in technology have led to increased accuracy and affordability of automatic transcription services, making them preferable over the laborious and time-consuming process of manual transcription. STR allows users to take advantage of speech-to-text transcription services to transcribe their own verbal response data. STR is partially tailored to child development researchers utilising the Codes for the Human Analysis of Transcripts (CHAT) though the code is generic enough to output unformatted transcriptions. STR allows transcription of single words and multi-speaker dialogues in 50+ languages. We describe STR, provide a tutorial for CHAT-formatted transcriptions, describe settings available for customising transcription and conduct a brief analysis of the efficiency and accuracy of transcription. Speech that was transcribed in over half an hour by trained transcribers was transcribed in less than two minutes (with ~90% accuracy) by Rev AI. Considering the additional time needed for error correction and CHAT formatting, we estimate that manual transcription takes twice as long as transcribing with assistance from STR.