UD-toolkit

UD-toolkit is an NLP toolkit built around UDPipe providing out-of-the-box language tools. Models from UD 2.0 are dynamically downloaded so that you can focus on the task at hand.

Note that while this software is licensed under the GPL, the UD 2.0 models are distributed under the CC BY-NC-SA license which explictly prohibits commercial use.

Installation

pip install ud-toolkit

Usage

To get started, load a model:

>>> from udtk import Model
>>> m = Model("english")

UD-toolkit will download the model for you. For a complete list of available model, see udtk.models.LANGUAGES.

To tokenize, lemmatize and get part-of-speech tags, ud-toolkit provides easy to use convenience functions.

>>> s = "Time flies like an arrow. Fruit flies like a banana."
>>> m.tokenize(s)
['Time', 'flies', 'like', 'an', 'arrow', '.', 'Fruit', 'flies', 'like', 'a', 'banana', '.']
>>> m.lemmatize(s)
['time', 'flie', 'like', 'a', 'arrow', '.', 'fruit', 'fly', 'like', 'a', 'banana', '.']
>>> m.pos_tag(s)
[('Time', 'NOUN'), ('flies', 'VERB'), ('like', 'ADP'), ('an', 'DET'), ('arrow', 'NOUN'), ('.', 'PUNCT'), ('Fruit', 'NOUN'), ('flies', 'VERB'), ('like', 'ADP'), ('a', 'DET'), ('banana', 'NOUN'), ('.', 'PUNCT')]

For more advanced usage, you can use Model.process():

>>> [(w.lemma, w.xpostag) for w in m.process(s, tag=True)]
[('time', 'NN'), ('flie', 'VBZ'), ('like', 'IN'), ('a', 'DT'), ('arrow', 'NN'), ('.', '.'), ('fruit', 'NN'), ('fly', 'VBZ'), ('like', 'IN'), ('a', 'DT'), ('banana', 'NN'), ('.', '.')]

Supported languages

• ancient_greek
• ancient_greek-proiel
• arabic
• basque
• belarusian
• bulgarian
• catalan
• chinese
• coptic
• croatian
• czech
• czech-cac
• czech-cltt
• danish
• dutch
• dutch-lassysmall
• english
• english-lines
• english-partut
• estonian
• finnish
• finnish-ftb
• french
• french-partut
• french-sequoia
• galician
• galician-treegal
• german
• gothic
• greek
• hebrew
• hindi
• hungarian
• indonesian
• irish
• italian
• japanese
• kazakh
• korean
• latin
• latin-ittb
• latin-proiel
• latvian
• lithuanian
• norwegian-bokmaal
• norwegian-nynorsk
• old_church_slavonic
• persian
• polish
• portuguese
• portuguese-br
• romanian
• russian
• russian-syntagrus
• sanskrit
• slovak
• slovenian
• slovenian-sst
• spanish
• spanish-ancora
• swedish
• swedish-lines
• tamil
• turkish
• ukrainian
• urdu
• uyghur
• vietnamese