Bitbucket is a code hosting site with unlimited public and private repositories. We're also free for small teams!

Close

opencorpora-tools

This package provides Python interface to http://opencorpora.org/

Installation

pip install opencorpora-tools

If you have python 2.6 then argparse and ordereddict packages are required:

pip install argparse
pip install ordereddict

Usage

Obtaining corpora

Opencorpora-tools works with XML from http://opencorpora.org/.

You can download and unpack the XML manually (from 'Downloads' page) or just use the provided command-line util:

$ opencorpora download

Run opencorpora download --help for more options.

Using corpora

Initialize:

>>> import opencorpora
>>> corpus = opencorpora.CorpusReader('annot.opcorpora.xml')

Get table of contents:

>>> corpus.catalog()
[('1', '"Частный корреспондент"'),
 ('2', '00021 Школа злословия'),
 ('3', '00022 Последнее восстание в Сеуле'),
 ('4', '00023 За кота - ответишь!'),
...

Work with documents:

>>> seoul_words = corpus.words('3')
>>> seoul_words
['«', 'Последнее', 'восстание', '»', 'в', 'Сеуле', ...

>>> corpus.documents(categories='Тема:ЧасКор:Книги*')
[Document: 21759 2001-2010-й: книги, которые потрясали,
 Document: 12824 86 снов, вызванных полётом пчелы вокруг граната за секунду до пробуждения,
 Document: 10930 А бойтесь единственно только того, кто скажет: «Я знаю, как надо!»,
 ...

opencorpora.Corpora is modelled after NLTK's CorpusReader interface; consult with http://nltk.googlecode.com/svn/trunk/doc/book/ch02.html to get an idea how to work with the API. It it not exactly the same, but should be very similar.

Performance

OpenCorpora XML is huge (>250MB) so building full DOM tree requires a lot of memory (several GB) and should be avoided.

opencorpora-tools handles it this way:

  1. corpus.get_document(doc_id) or corpus.documents(doc_ids) don't load the original XML to memory and don't parse the whole XML. They use precomputed offset information to slice the XML instead. The offset information is computed on first access and saved to "<name>.~" file.

    Consider document loading O(1) regarding full XML size. Individual documents are not huge so they and loaded and parsed as usual.

  2. There are iterator methods for all corpora API (corpus.iter_words, etc).

Development

Development happens at github and bitbucket:

The main issue tracker is at github.

Feel free to submit ideas, bugs, pull requests (git or hg) or regular patches.

Running tests

Make sure tox is installed and run

$ tox

from the source checkout. Tests should pass under python 2.6..3.3 and pypy > 1.9.

Recent activity

Mikhail Korobov

Mikhail Korobov pushed 1 commit to kmike/opencorpora-tools

ac6a8e7 - breaking change: make tagged_... methods compatible with NLTK. Old methods (that return normalized word and all word parses) are renamed to parsed_...
Tip: Filter by directory path e.g. /media app.js to search for public/media/app.js.
Tip: Use camelCasing e.g. ProjME to search for ProjectModifiedEvent.java.
Tip: Filter by extension type e.g. /repo .js to search for all .js files in the /repo directory.
Tip: Separate your search with spaces e.g. /ssh pom.xml to search for src/ssh/pom.xml.
Tip: Use ↑ and ↓ arrow keys to navigate and return to view the file.
Tip: You can also navigate files with Ctrl+j (next) and Ctrl+k (previous) and view the file with Ctrl+o.
Tip: You can also navigate files with Alt+j (next) and Alt+k (previous) and view the file with Alt+o.