Skip to main content

Word sense disambiguation library

Project description

This repository contains scripts and expriments related to the Sense frequencies project, and an rlwsd python package for WSD (word sense disambiguation) for Russian language.

rlwsd package

This package can perform WSD for Russian nouns described in the of Active Dictionary of Russian (currently, only the first volume is published with letters “А” - “Г”).

Installation

The package currently works only on CPython 3.4+. Install with pip:

pip3 install rlwsd

The package requires models that are not hosted on PyPI and most be downloaded separately (about 2.3 Gb total):

python3 -m rlwsd.download

Models are re-downloaded even if they are already present. In case of problems (download does not finish, etc.) you can download models manually from rlwsd.download.MODELS_URL and extract them into the models folder inside rlwsd (package) folder.

Usage

Most functionality is provided by the model class. Model for each word must be loaded separately:

>>> import rlwsd
>>> model = rlwsd.SphericalModel.load('альбом')
>>> model.senses
{'1': {'meaning': 'Вещь в виде большой тетради ...',
       'name': 'альбом 1'},
 '2': {'meaning': 'Книга тематически связанных изобразительных материалов ...',
       'name': 'альбом 2.1'},
 '3': {'meaning': 'Собрание музыкальных произведений ...',
       'name': 'альбом 2.2'}}
>>> model.disambiguate('она задумчиво листала', 'альбом', 'с фотографиями')
'2'

You can also get a list of all words with models:

>>> import rlwsd
>>> rlwsd.list_words()
['абрикос',
 'абсурд',
 'авангард',
 ...
 'гусь',
 'гуща']

A large word2vec model is used internally. By default it is loaded once, one the first call to .disambiguate method, which takes noticeable time. There is an option to load word2vec model in a separate process by running w2v-server command, which starts a server, and exporting W2VSRV environment variable with any non-empty value:

# in the first terminal window
$ w2v-server
running...
# in the second terminal window
$ export W2VSRV=yes
$ python

In this way you can leave the w2v-server running and save time on word2vec model reloads.

License

License is MIT

Project details


Release history Release notifications

This version
History Node

0.1.2

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
rlwsd-0.1.2-py2.py3-none-any.whl (13.1 kB) Copy SHA256 hash SHA256 Wheel 3.4 May 22, 2016
rlwsd-0.1.2.tar.gz (11.0 kB) Copy SHA256 hash SHA256 Source None May 22, 2016

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page