Polly 2
I am currently contributing to the Polly 2 project (Part 1 : Time Series-Speaker Turns Labelling & Part 2 : Transcription) ms_Malay with OneForma. In Time Series, I first need to review audios and their pre-annotated Speaker Turns, Gender, and Timestamp interval. Each timestamp needs to start and end within a 500-1000 milliseconds buffer. If there is any speech that was not detected by AI, I need to create a new timestamp and label it with the correct speaker turn number and gender. Once everything is reviewed and verified, the task will be submitted. For Transcription, I have to review the audio for any clarity or language issues, especially if there are multiple speakers involved. I have to then review and edit the pre-transcribed text according to the locale language guideline. Only the primary speaker's speech should be transcribed. Once done, task is submitted.