I started an annotation task with multiple annotators with :
$prodigy ner.manual mydata blank:fr source.jsonl --label A,B
I then exported the DB mydata with :
$prodigy db-out mydata > export.jsonl
I edited export.jsonl, resulting with edit.jsonl
(minor changes in the "_annotator_id" and "_session_id" fields such as “mydata-léa” –> “newdata-lea”)
I now want to create a new DB from edit.jsonl, so that each annotator can continue the job from where they left it. My issue is that after the following command lines, the annotators have to start from the first example again :
$prodigy db-in newdata edit.jsonl
$prodigy ner.manual newdata blank:fr source.jsonl --label A,B
I do have set “exlude_by” to “input”in the conf file though.
Do you have any ideas ?
Hi! If the only changes are in the underscore attributes and don't touch the actual input data (e.g. "text"), your workflow should definitely work Some questions to debug this:
When the first example repeats, can you log the JSON and see if the _task_hash and _input_hash values match those of the same example in your exported data? (Tip: If you're on the latest version of Prodigy, you should be able to do it in your browser's JavaScript console by typing window.prodigy.content!)
Since you're using multi-user sessions, are you accessing Prodigy with one of the sessions? Maybe you're accidentally starting it up with no session ID or a new session, so your examples are repeated again from the beginning instead of excluding those already annotated in the session?