Release 0.7.1

• Works on Python 3.7 again

Evaluation scores:
TAG: 95.17
POS: 94.76
MORPH: 65.30
LEMMA: 93.35
UAS: 85.08
LAS: 79.82

License: MIT

Release 0.7.0

• Compatibility with spaCy v3.1
• Minor improvements to analysis: prefer non-compound words

Evaluation scores:
TAG: 95.17
POS: 94.76
MORPH: 65.30
LEMMA: 93.35
UAS: 85.08
LAS: 79.82

License: MIT

Release 0.6.0

• Improved tagging and parsing accuracy by pretraining
• Improved lemmatization accuracy by better handling of ambiguous inflections
• Morphological features (case, verb tense, person, etc.)
• Properly set POS SPACE on whitespace tokens

Evaluation scores:
TAG: 95.33
POS: 94.85
MORPH: 66.57
LEMMA: 93.40
UAS: 84.85
LAS: 79.17

License: MIT

Release 0.5.0

• Ported to spaCy 3.0. Does not support SpaCy 2.x anymore.

Evaluation scores:

TAG: 94.55
POS: 94.08
LEMMA: 93.26
UAS: 81.46
LAS: 75.45

License: MIT

Release 0.4.1

• Published as a PyPI package. The package name is spacy_fi_experimental_web_md

License: GPL v3.0

Release 0.4.0

• Ported to SpaCy 2.3
• Include 500k keys and 20k vectors like in the official *_md models
• Include the word vectors for the most frequent words

License: GPL v3.0

Release 0.3.0

• Extract noun phrases
• Lemmatize conjugated abbreviations: EU:ssa => EU
• Requires SpaCy 2.2.4 or later

License: GPL v3.0

Release 0.2.0

• Tagging auxiliary verbs as AUX (previously VERB) following the UD convention
• Fixed bugs in lemmatization of compounds words: ilmakuivata, esiopetus, etc
• Improved lemmatization of pronouns, especially clitics: sinäkin, mekään, etc
• Using the same Finnish tokenizer rules as the spaCy master branch

License: GPL v3.0

Version 0.1.0

License: GPLv3