indralab/adeft: 0.7.0

This release updates the longform expansion discover algorithm in AdeftMiner to combine the Acromine based approach with an alignment based scorer that we have developed. Alignment based scoring algorithms look for common subsequences between the shortform and longform candidates, different approaches scoring matches in a variety of ways. We have combined the two approaches by taking weighted averages of normalized Acromine scores and alignment based scores for each longform candidate, with the weight assigned to the alignment based score increasing for rare expansions.

The AdeftRecognizer has also been updated to allow the raw longform expansion to be recovered as it actually appears in a text. Previously only a normalized expansion was recovered.