Web interface for ASR transcriber

Abstract

With the rise of artificial intelligence and machine learning, speech-to-text (STT) technology has become more accurate and precise than ever. Since its conception, STT has proven to have a huge potential in many fields, one of them obviously being transcribing. Transcribing is an act of making a written copy of the content of an audio or a video. Even now, transcribing is still mostly a manual and laborious job, which can be extremely cost-inefficient, time-consuming and prone to human error. By leveraging the powerful ability of STT, one can significantly cut down on time loss and improve upon the accuracy and efficiency of the process. However, creating an STT model requires a substantial amount of data, both in terms of audio/video and transcription. The data then needs to be tracked, managed, and corrected accordingly so that it can be used to train the model. Current process if done manually has proven to be highly inefficient and time-consuming. Thus, there is a need for a centralized interface that can both allow users to interact and manage data, as well as connect said data to the STT model. The Transcriptor was created as the solution to the problem. As a web application, the Transcriptor offers features like user management, upload/download transcriptions and built-in transcription editor. Initial testing shows that Transcriptor helps cut down on the time used to manage the data as well as increases productivity on manual correction of transcripts.