The dataset contains list of agriculture videos, corresponding questions that can be asked and the answer that can be generated from the videos.
Further in the columns, the dataset gives what is the source of information which could be:
1) Transcript
2) Object
3) Scene description
4) Text over lay
The next columns have timestamps.

The videos are publicly available on youtube and following languages are covered:
1) English
2) Bhojpuri
3) Hindi
4) Odiya
5) Telugu

The dataset can be used for training models to generate answers from the videos based on what is spoken as well as what is shown.