UI request: "flag" button

I have a feature/clutter request for the UI. Would it be possible to have a “flag” button somewhere in the annotation interface, that would add a flag or mark to the example in the DB? It could be very small and subtle and out of the way. I’ve wanted this a few times so I could go mark examples to go back and check later for situations like:

  • how did this Spanish text end up in my English stream?
  • wow, that’s a weird pattern/phrasing. I want to look at that more
  • this document is very substantively interesting and I want to look it up later for some qualitative analysis
  • this is a multiword NER that needs manual annotation

Right now I’ve just been copying down the document ID from the meta field, but that slows things down a lot. I wanted to point this out as something that I would find useful, but I know you’re fighting the good fight against more buttons and clutter.

1 Like

Yes, that’s a nice idea! :+1: Something similar came up in a thread a while ago and I was thinking about a “bookmark” button, for example, in the top right corner of the annotation card. But I think I like the “flag” terminology even better – we could then add "flagged": true to the example, and you’d be able to use db-out with an option --flagged-only to export only the flagged examples.

1 Like

I love it! And there could be a config option to display the button or not, which could be useful if you don’t want annotators getting confused by it.

Quick update: The upcoming version will come with experimental support for flagging tasks that can be enabled by setting "show_flag": true. This will add a little flag icon to the top right corner of the annotation card, which can be toggled via click or by pressing F.


Flagging a task will simply add "flagged": true. When exporting annotations with db-out, you can set --flagged-only to only export flagged tasks.

2 Likes

Any command to directly use it in script ?

The option is not set on the command line, so there’s nothing you should have to do if you’re running Prodigy via a Python script. Just add "show_flag": true to your prodigy.json, or add 'show_flag': True to the 'config' option returned by your recipe (if you’re using a custom one and only want to set this option for one particular recipe).