- Separeted
udify
andelmo
pipelines to camphr_allennlp - now camphr doesn't depend on allennlp.
- [dependencies] Bump scikit-learn from 0.22.2.post1 to 0.23.1 #61
- [dependencies] Bump pytest from 5.3.2 to 5.4.3 #60
- [closed] support allennlp v1 #59
- [closed] Improvement for サ変 of KNP #56
- [closed] refactor #55
- [bug] fix transformers eval batchsize failure #50
- [bug] Proper treatment of PUNCTs for KNP #48
- [enhancement] dependency improvement for KNP #47
- [enhancement] update transformers dependency #46
- [CI] Skip slow ci if unnecessary #45
- [enhancement] Refactor/knp dependency parser #44
- [enhancement] Tentative dependencies for KNP #43
- [enhancement] juman TAG_MAP tentative support #41
- [bug] Fix misuse
Vocab()
in Language instantiation #42
- [enhancement] Revert sentencepiece lang from v0.4 #40
- [enhancement] add
functools.lru_cache
to knp extensions. #39
No changelog for this release.
- [enhancement] tag and bunsetsu can be directly got from token #38
- [enhancement] Feature/knp para noun chunks #37
- [bug] fix noun chunker for para phrase #36
- [enhancement][refactor] Refactor/knp noun chunker #35
- Separate parallel clause in noun chunks into two or more chunks #34
- Support knp noun chunker and knp dependency parser #33
- It is now possible to retrieve KNP result from
spacy.doc
(#31)
Removed the version restriction python<3.8
. This will allow users to install camphr with python3.8
, but macos users will fail. see (#29) for details.
- juman and knp now accepts longer text (#23)
- fix transformers requirements (#24)
- Multi labels textcat pipe for transformers (#14)
- fix
camphr.utils.get_requirements_line
- Computing val loss in TorchLanguage.evaluate` #13