How do we inspect dataset sessions?

, ,
1

I can list the sessions by:

prodigy stats DATASET_NAME -ls

And then drop the session by:

prodigy drop SESSION_ID

But how do we inspect the individual sessions? I’d like to see ACCEPT / REJECT counts, inspect individual labels, etc.

There was a case where we think one annotator did a REJECT on all of their samples and we’d like to investigate.

Thanks!

2

A session is really just a “special” dataset and its name is the session ID – so you can use Prodigy’s regular database commands to export or output the annotations. For example:

prodigy db-out SESSION_ID | less
prodigy db-out SESSION_ID > session.jsonl

You can also use the print recipes to get nicely formatted command line output:

prodigy ner.print-dataset SESSION_ID | less -R
prodigy textcat.print-dataset SESSION_ID | less -R

(We’re hoping to improve some of the dataset and session handling as part of the corpus management update. I’ve also been thinking about writing a little recipe that makes it easier to navigate the individual sessions. They’re all timestamped, so it won’t be difficult to group them together by year, month, week, day etc.)

3

Oh thanks. I should’ve tried that first :slight_smile:

4

Also is there a way to only show sessions from one dataset?