Heuristic function that specifically says a span is NOT an entity? about skweak HOT 3 CLOSED

+1, exactly the same question :)

from skweak.

Sorry for the answering delay! But I think I have a relatively simple solution to your problem: the easiest is to create a normal heuristic function (like with FunctionAnnotator), and have it output a special label such as NOT_A_LABEL . Then, when aggregating all the labels with the HMM or the MajorityVoter, specify that NOT_A_LABEL is an underspecified label of O:

hmm = skweak.aggregation.HMM("hmm", ["PERSON", "DATE", "MONEY"])
hmm.add_underspecified_label("NOT_A_LABEL", ["O"])

This way, the aggregation will know that your labelling function has produced a signal that is equivalent to "O" -- but since the label is defined as an underspecified label (and not a "concrete" label), it will also know that NOT_A_LABEL is not a "real" label.

I've had to make some minor changes to the code to make it working, let me know whether this works!

from skweak.

Thanks for the reply! I'll try it out and let you know! Thanks!

from skweak.