Vāgyojaka: An Annotation and Post-Editing Tool for Automatic Speech Recognition

Vāgyojaka is an open-source post-editing and annotation tool for automatic speech recognition (ASR) that aims to reduce the human effort required to correct the ASR results. We adopt a dictionary-based lookup method to highlight the incorrect words in the ASR transcript and give suggestions by generating the closest valid words. For curating the speech corpus, we provide a rich list of tagset that captures various spoken audio features. Further, we conducted a user study to evaluate the effectiveness of our tool and observed that post-editing requires 1/3 lesser time than editing without using our tool. The user study can be found on our website.