Closed bruffridge closed 3 years ago
Looks like MATCH's preprocessing scripts expect the json files to come with a 'label' field, but you can omit the labels file they produce in testing.
For consistency, my recommendation would be to include a 'label' field in the json objects, but it can be an empty array.
So to generate a prediction for an unlabelled paper, you invoke the model with json formatted as described here? https://github.com/nasa-petal/PeTaL-labeller/issues/43
Yes, that format should do.
Is it the same as the format for the training data just without the
label
field? See #43 . This issue blocks #45.