Respect encoding when reading binary keyed vectors #3309
Merged
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
Current implementation fails to read keyed vectors that have iso-8859-1 encoding in the words when encoded in binary format. An example of this type of a file can be seen in the turkuNLP finnish embeddings: http://dl.turkunlp.org/finnish-embeddings/finnish_s24_skgram.bin This file is quite trivial to load by passing the encoding to the vector loading function. It is also logical that when user asks KeyedVectors.load_word2vec_format(filename, binary=True, encoding='iso-8859-1') The library would try to load the file assuming that the matrix is in binary format and the words are encoded using iso-8859-1 encoding.
|
Makes sense, thanks. If this works and all tests pass, let's include this PR in the upcoming release @mpenkov . |
Codecov Report
@@ Coverage Diff @@
## develop #3309 +/- ##
===========================================
- Coverage 81.43% 79.53% -1.90%
===========================================
Files 122 68 -54
Lines 21052 11783 -9269
===========================================
- Hits 17144 9372 -7772
+ Misses 3908 2411 -1497
Continue to review full report at Codecov.
|
|
One test failed, can you check please @alhoo? |
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Outdated suggestions cannot be applied.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.
Current implementation fails to read keyed vectors that have iso-8859-1
encoding in the words when encoded in binary format. An example of this
type of a file can be seen in the turkuNLP finnish embeddings:
http://dl.turkunlp.org/finnish-embeddings/finnish_s24_skgram.bin
This file is quite trivial to load by passing the encoding to the vector
loading function. It is also logical that when user asks
KeyedVectors.load_word2vec_format(filename, binary=True, encoding='iso-8859-1')
The library would try to load the file assuming that the matrix is in
binary format and the words are encoded using iso-8859-1 encoding.