Levenshtein distance and fuzzy matching#
scikitplot.levenshtein provides one small, stable facade for exact
unit-cost Levenshtein edit distance, normalized similarity, deterministic
ranking and optional Corpus retrieval.
The base behavior never depends on a third-party package. Optional accelerators are discovered lazily.
Distance, normalized similarity and the first ranked match.
Metric semantics, fallback boundaries and provenance.
Internal, RapidFuzz, explicit GPL backend and pure Python.
Stable ties, limit, key and score_cutoff.
Use normalized edit similarity as a retrieval scorer.
Complexity, streaming choices and what cutoff does today.
Scope#
The facade intentionally does not try to expose every feature from
RapidFuzz or the external Levenshtein project. It keeps one portable metric
contract and lets optional implementations accelerate it.
For transpositions, weighted operations, edit scripts, median strings or a broader fuzzy-matching toolbox, use a specialized library directly.
Examples#
See Levenshtein for executable, network-free examples.