I am running a ner.make-gold recipe in combination with a pretrained model to help with the already learned Labels. I am also using a custom theme (with altered colours of the labels highlights to make annotation even faster as seeing colors is faster than reading the labels )
Now when I tried to add the flagging option in the PRODIGY.JSON, it doesnt seem to show in the UI (nor does the shortcut ‘f’ works). I have tried to put “show_flag”: true in different places, but no luck.
Is it possible to have the flagging option with ner.make-gold while using a custom pretrained model to suggest labels during annotation?
Hi! I think there might be a small typo in your JSON config: you’ve added the "show_flag": true in the "custom_theme" object, which is only really for theme/color customisations. To be recognised by Prodigy, the setting needs to be at the root level. Could you edit your config and try again?
p.s. and thanks for the hint for handy json formatter in posts
p.p.s I also tried to force flagging through the ner.make-gold recipe, by editing the original recipe and adding “show_flag”: True in the config return part of the recipe, but didnt worked as well.
Do you have any other prodigy.json or .prodigy.json config files that set "show_flag": false, e.g. in your home directory or your project directory?
Your local config in your working directory will overwrite the global config, and that combined config will then overwrite the pre-defined recipe config. The idea behind it is that you can set custom personal preferences on your machine (like, a theme), project-specific settings (like, the batch size or UI behaviour) for the current directory, and the more generic defaults in the recipes themselves.
Do you have any other prodigy.json or .prodigy.json config files that set "show_flag": false , e.g. in your home directory or your project directory?
Just searched my environment, and this is the only prodigy.json file, it is located in C:\Users\ivo\.prodigy
so seems no override.
However just a guess.. since the manual recipe contains a div which holds the label names (the top purple block). Perhaps the flag button is below them in the html? Although the shortcut 'f' should have worked.
Thanks for checking! And hmm, this is definitely strange
Yeah, I was wondering this, too. It might be related to a small fix I made to make the header sticky for very large tasks. I'll need to investigate this and try to reproduce it. At least, if the suspicion is correct, I should be able to give you one small line to add to the "global_css" setting that fixes it.
So you pressed f, saved and then checked the examples in the dataset and they didn't have "flagged": true? If so, then yes, that's confusing, yes...
Ohh, ok – the f should only set "flagged": true on the task, and not do anything else. Sorry if this was confusing or misleading. In addition, you’d still have to accept/reject/ignore the task. This allows you to flag any task, regardless of the answer (e.g. to investigate it further, export it as a nice example for your annotation manual or to show your colleague etc.)
The visual indicator in the UI is that the flag icon turns red – but in this case, it’s not happening because the flag icon is covered by the title bar.
Ok, this was definitely the explanation then. I found the problem and already fixed it for the next release. In the meantime, try adding the following at the root of your prodigy.json:
That’s very mysterious! I really don’t see how this wouldn’t work
Could you do me a favour and go into your browser’s developer console, right-click on <html> and copy the markup of the whole page to a .html file? (In Firefox, this should be called “Edit as HTML”).
It’ll be quite large, but you should be able to drag and drop the file into the reply text field here, just like images. Alternatively, you can also email it over.
