10x Faster Levenshtein Distances #3233
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This PR replaces the native Python implementation of the Levenshtein distance with a SIMD-accelerated version from the StringZilla library. On ~5 letter words from a typical English corpus - StringZilla is over 10x faster than current Python implementation in NLTK. On a multilingual corpus with longer words - same result.
tomaarsen
reviewed
Feb 19, 2024
|
[CI: retest] |
ashvardanian
force-pushed
the
optimize/levenshtein
branch
from
61c407d
to
77c7712
Compare
|
[CI: retest] |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR replaces the native Python implementation of the Levenshtein distance with a SIMD-accelerated version from the StringZilla library.
On ~5 letter words from a typical English corpus - StringZilla is over 10x faster than current Python implementation in NLTK (1.24 s ± 57.4 ms vs 13.5 s ± 287 ms per loop across 7 runs). On a multilingual corpus with longer words - same result.
ifbranches than most native implementations, so it's faster than many implementations in the serial mode. When AVX-512 is available it can use specialized assembly instructions to accelerate evaluation.