fix: correct reference length calculation #195
Closed
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Outdated suggestions cannot be applied.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.
Summary
This PR fixes the way brevity penalty (specifically the effective reference corpus length) is calculated in BLEU.
Previously,
len_reference
was calculated asmin([len(ref) for ref in references_tokenized])
. However, this is incorrect, because according to the paper, we need to find the "best match length", not the minimum reference length.For more information, see wikipedia - brevity penalty and nltk implementation.
Test plan
I added another unit test to
test_bleu.py
and compared the results of the calculations to the results of thenltk.translate.bleu_score.corpus_bleu
function to make sure the implementation is correct.