Size of the dataset

Can anyone suggest what would be an ideal/bare minimum data size to start the training with? I understand that is heavily dependent on the variety of format we handle but I tried with some 100 docs and didn't get any success on any field. Not even close. So If we have some knowledge on the numbers one can plan accordingly.

@rrajp you need to at least need a set of 500-800 docs to train.

do i have to anotate to get the JSON format manualy for 800 docs ? @janhavisawal

do i have to anotate to get the JSON format manualy for 800 docs ? @janhavisawal

Yes, you need too. You can use makesense.ai for that.