I have a task of ~30k annotations. Around 9k of those I already know the answer. I’d like to import these answers into my db. Is that possible?
Yes, if you convert them to Prodigy’s JSON format, you can use the db-in
command to import them to a dataset
But what fields are needed? I only have the answers and the meta
fields but not task_id
etc. Will those be created on import?
Yes, the required internals like the hashes will be set automatically if they’re not in the data yet. So the task really only needs the data (text, label, spans, meta, whatever) and the "answer"
. If no answer is present, it can also be added automatically – it defaults to "accept"
, but you can customise it by setting the --answer
option. You can also set --dry
to do a dry run.