I just saw the prodigy1.7 update. It’s really great for multiuser annotations.
I wanted to know that is there any way to separate out the data from each user session.
Yes, each user session will be created as a session dataset (just like the timestamp session datasets). So if your dataset name is
ner_annotations and your session is
akshita, the session dataset created in the database will be
ner_annotations-akshita. You can then use that dataset name for training or export it using
Oh! That’s great. And does it store all the session names?
In case I am not sure of the sessions that were started in between?
I’m not 100% sure I understand your question correctly. If you access Prodigy with the identifier
/?session=akshita, those annotations will be added to the regular dataset, e.g.
ner_annotations, and the session dataset
ner_annotations-akshita. If you access Prodigy again with the same session name later, your annotations will be assigned to the same session again. If someone else accesses Prodigy with a different session, their annotations will also be added to the main dataset and their session dataset as well. So in the end, you might have one main dataset
ner_annotations that bundles all annotations, and separate session datasets that only contain the annotations of the named sessions. For example,
ner_annotations-alex, and so on.
To see an overview of all available sessions, you can run
prodigy stats -ls.
If you check out your
PRODIGY_README.html, there’s also a section on the named multi-user sessions with an example of this.
That answered my question. Thanks a lot.
I was basically looking for the session datasets, which I think can be fetched using ‘db.sessions’.
And this update is great!!! A solution to so many problems.