Vinsingh's picture
Create README.md
581023d verified
|
raw
history blame
716 Bytes
metadata
license: apache-2.0
task_categories:
  - translation
  - automatic-speech-recognition
language:
  - hi
  - or
tags:
  - Agriculture
  - video
size_categories:
  - 1K<n<10K

This dataset consists of agriculture videos in hindi and oriya. The dataset consists of mulitple xls files and each xls file has column of video urls (youtube video links) and corresponding transcripts.

The transcripts are:

  1. Generated by ASR models (for the purpose of benchmarking)
  2. Manual transcripts
  3. Time stamps
  4. Manual translations

This dataset can be used for training and benchmarking domain specific models for ASR and translation. The time stamps serve as the soruce of audio and the corresponding human transcripts can be used.