AKAI Answer quality: STT enhancements

[MaunicaKolla815]

Context

Currently Fairseq is used for STT (odia) in AKAI. We need to verify if BHASHINI Model's latency against the existing

Requirements

• [x] Dataset with audio STT queries (message ids) - @MaunicaKolla815
• [ ] Re-run the queries with Bhashini model - @Amruth-Vamshi

cc: @GautamR-Samagra

[MaunicaKolla815]

https://docs.google.com/spreadsheets/d/1jbj5mqmDhsIaJg2i3yRSXiNXzNy4VR_UiSJ_DWBZ2tg/edit?usp=sharing - Dataset for STT queries (july + august)