I'd like to create a recipe to do multilabel classification. I already have one label annotated, and I'd like to add more if they apply. I was thinking of creating a recipe where the one label annotated is not shown or selected by default, while the other labels can show up as empty checkboxes. Can you please help me out with some pointers as to how I can go about writing this?
Hi! This should be pretty straightforward to do in a custom recipe similar to this one:
Prodigy's input and output data formats are the same and you can feed it already annotated examples that will then be displayed in the UI. The choice interface stores the IDs of the accepted answers as the key "accept" – so all you'd have to do is feed in data that specifies the selected labels, e.g. "accept": ["SOME_LABEL"]. See here for the full data format of the choice interface: https://prodi.gy/docs/api-interfaces#choice