TEXTA Tagger

Installation

Using built package

pip install texta-tagger

Using Git (for development)

pip install git+https://git.texta.ee/texta/texta-tagger.git

Usage

Predicting Using Zipped Model

Predicting without MLP & lemmatization

>>> from texta_tagger.tagger import Tagger
>>>
>>> t = Tagger()
>>> t.load_zip('test_data/tagger.zip')
True
>>> print('Tagger:', t)
Tagger: Eesti
>>> t.tag_text('eesti keel ja eesti meel')
{'prediction': True, 'probability': 0.9999999322365133}

Predicting with MLP & lemmatization

Predicting with lemmatization requires either a running MLP server or MLP Python package installed. In following example MLP from package is used:

>>> from texta_tagger.tagger import Tagger
>>> from texta_mlp.mlp import MLP
>>>
>>> mlp = MLP()
>>>
>>> t = Tagger(mlp=mlp)
>>> t.load_zip('test_data/tagger.zip')
True
>>> print('Tagger:', t)
Tagger: Eesti
>>> t.tag_text('eesti keel ja eesti meel')
{'prediction': True, 'probability': 0.9999999322365133}

In following example MLP server version is used:

>>> from texta_tagger.tagger import Tagger
>>> from texta_tagger.mlp_analyzer import get_mlp_analyzer
>>>
>>> mlp = get_mlp_analyzer(mlp_host="http://my-mlp-server:5000")
>>>
>>> t = Tagger(mlp=mlp)
>>> t.load_zip('test_data/tagger.zip')
True
>>> print('Tagger:', t)
Tagger: Eesti
>>> t.tag_text('eesti keel ja eesti meel')
{'prediction': True, 'probability': 0.9999999322365133}

Training

TODO

Environment Variables

• TEXTA_TAGGER_MLP_URL - MLP host used for lemmatization (e.g. http://mlp-dev.texta.ee:5000)
• TEXTA_TAGGER_MLP_MAJOR_VERSION - MLP major version (2/3).