Odd results with few values?

[axiak]

Odd results with few values?

[axiak]

Hey -- sorry for the late comment. By default we're using edit distance by the well-known Levenshtein algorithm to determine similarity. As you've noted there are times when it seems odd. You can specify use_levenshtein=False to get cosine similarity based on matching trigrams/bigrams. Here's an example of that:

>>> f = FuzzySet(use_levenshtein=False)
>>> f.add(u'Conor Hedley')
>>> print(f.get(u'directory manager'))
[(0.06537204504606135, 'Conor Hedley')]
>>> f.add(u'man')
>>> print(f.get(u'directory manager'))
[(0.140028008402801, 'man')]