Expand description
Python’s difflib — Ratcliff/Obershelp similarity and
get_close_matches.
The scanner’s near-miss cutoff (0.72) is a behavioural boundary, so the
ratio has to agree with CPython to the last digit rather than merely be
“a similarity measure”. This is a direct transcription of
SequenceMatcher, not an equivalent algorithm.
Autojunk is on, as it is by default in CPython: once b reaches 200
elements, any element occurring more than len(b) // 100 + 1 times is
dropped from the index. That rarely touches a scanner tag but decides
the score of a long line of dialogue (0.295 with it, 0.873 without, for
one pair of 220-character sentences).
Structs§
- Sequence
Matcher difflib.SequenceMatcher(None, a, b)over character sequences.
Functions§
- get_
close_ matches difflib.get_close_matches(word, possibilities, n, cutoff).