-
Notifications
You must be signed in to change notification settings - Fork 8
Description
Currently the score for each of the fields used in an "hypothesis" when compared with candidates retrieved from Solr records, is added to the list of "evidences". The final score is the sum of these "evidences". In the current implementation, the scores for these individual fields seem to be taken "as is", i.e. a score for the author field weighs just as heavily as a score for the publication year. It seems that we should be considering whether to assign weights to these individual scores.
The logical location to implement this seems to be the add_evidence method which has the numerical value for the evidence and the associated label. The weights should follow from the value of this label.
Another possible location is the get_score function in this part
if self.score is None:
self.score = sum(self.evidences)
Instead of just summing over the list with evidences, each entry in this list should be multiplied with the appropriate weight (based on the values in self.labels)