Fix #4676: Prevent gibberish detector from flagging legitimate copyrights #4678
+31
−7
Conversation
This file contains hidden or 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
NullPointer-cell
force-pushed
the
fix-4676-copyright-detection-regression
branch
5 times, most recently
from
c493323 to
0aa57a9
Compare
NullPointer-cell
force-pushed
the
fix-4676-copyright-detection-regression
branch
4 times, most recently
from
dac6930 to
c6e4e6c
Compare
added 2 commits
January 15, 2026 00:19
…trings - Skip gibberish detection for short lines (< 40 chars) with copyright indicators - Comprehensive copyright indicator list prevents false positives - Add training examples to good.txt for edge cases - Lenient assertion: handle overlapping probabilities during training - Fixes regression while preventing license detection false negatives Signed-off-by: Jayant Saxena <[email protected]> Signed-off-by: NullPointer-cell <[email protected]>
Signed-off-by: NullPointer-cell <[email protected]>
NullPointer-cell
force-pushed
the
fix-4676-copyright-detection-regression
branch
from
c6e4e6c to
65fdc52
Compare
JonoYang
requested changes
Jan 15, 2026
|
|
||
| # And pick a threshold halfway between the worst good and best bad inputs. | ||
| thresh = (min(good_probs) + max(bad_probs)) / 2 | ||
| if min(good_probs) > max(bad_probs): |
Member
There was a problem hiding this comment.
@NullPointer-cell why are the comments removed and why do we default to thresh = max(bad_probs) + 0.01
| def detect_gibberish(self, text): | ||
| text = ''.join(self.normalize(text)) | ||
| return self.avg_transition_prob(text, self.mat) < self.thresh | ||
| COPYRIGHT_INDICATORS = ( |
Member
There was a problem hiding this comment.
The logic to perform normalization of copyrights should be done in self.normalize.
| return self.avg_transition_prob(text, self.mat) < self.thresh | ||
| COPYRIGHT_INDICATORS = ( | ||
| 'copyright', '(c)', 'c)', '©', '@copyright', | ||
| 'author:', 'commit', 'portions:', 'rights reserved', |
Member
There was a problem hiding this comment.
These replacements are way too specific. I don't think we should add code that is only used to ensure we pass a specific arbitrary test.
|
|
||
| text_normalized = ''.join(self.normalize(text)) | ||
|
|
||
| if len(text_normalized) <= 4: |
Member
There was a problem hiding this comment.
what is the purpose of saying a string is gibberish if it is 4 characters or less in length?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Fixes #4676
Context
This PR supersedes #4677. The previous PR was automatically closed when the target branch
2402-detect-gibberish-copyrightwas deleted. I have rebased the fix ontodevelopas requested.Problem
Gibberish detection was incorrectly flagging legitimate copyright strings as gibberish, causing them to not be detected. This affected:
c) INRIA-ENPC.)@Copyright)commit ... Author:)Solution
Modified the gibberish detector to:
copyright,(c),©,@copyright,author:,commit)Tasks
Signed-off-by: Jayant Saxena [email protected]