Eric Kafe
Eric Kafe
Fix #3248 using file locks. In #3248, @naktinis describes a situation where users launch multiple useless parallel downloads of the same file, and cause race conditions. Rather than encouraging this...
Now that alternative data packages are available for all the pickles, the question arises: what to do with the old packages? Simply removing them seems very unsafe for those users...
One additional bug that came up while investigating #3345 and its CCG lexicon, was that variables that should have been alpha-converted were collapsed instead, resulting in unacceptable outputs that contained...
The latest version of this PR fixes issue #3473 for all the unit tests under _nltk/test/unit_, which can then load their datafiles from zipped packages.. The first commits solved the...
Four out of five nltk_data packages are set to unzip by default, although it seems likely that most package readers would only require minimal adjustments in order to not need...
This is a follow up to #3454, intended to find out why LazyCorpusLoader._unload() segfaults when testing with Python 3.13, and eventually fix it.
PR #3426 introduced an automated release workflow using _release.yml_. But pushing several release tags to my personal nltk clone (https://github.com/ekaf/nltk) has failed to produce an available experimental release. All failed...
Fix #3403: this pull request addresses issue #3403 by making the language name lookup functions in langnames.py more robust. It prevents potential KeyError exceptions by ensuring that missing language codes...
#### Motivation The current NLTK CCG implementation, while a strong and clear example of the formalism, has limitations in its linguistic expressiveness. It successfully demonstrates the core principles of syntactic...
## Problem The current implementation of `lang2q` (and the underlying `tag2q`) in `nltk.langnames` raises a `KeyError` if the BCP-47 tag is not found in the `bcp47.wiki_q` dictionary. This causes downstream...