fostroll / toxine

Tiny preprocessor for Russian text

RuMor: Russian Morphology project

Toxine: a tiny python NLP library for Russian text preprocessing

A part of RuMor project. It contains pipeline for preprocessing and tokenization texts in Russian. Also, it includes preliminary entity tagging. Highlights are:

• Extracting emojis, emails, dates, phones, urls, html/xml fragments etc.
• Tagging/removing tokens with unallowed symbols
• Normalizing punctuation
• Tokenization (via NLTK)
• Russan Wikipedia tokenizer
• brat annotations support

Installation

pip

Toxine supports Python 3.5 or later. To install it via pip, run:

$ pip install toxine

If you currently have a previous version of Toxine installed, use:

$ pip install toxine -U

From Source

Alternatively, you can also install Toxine from source of this git repository:

$ git clone https://github.com/fostroll/toxine.git
$ cd toxine
$ pip install -e .

This gives you access to examples that are not included to the PyPI package.

Setup

Toxine uses NLTK with punkt data downloaded. If you didn't do it yet, start Python interpreter and execute:

>>> import nltk
>>> nltk.download('punkt')

NB: If you plan to use methods for brat annotations renewal, you need to install the python-Levenshtein library. See more on the brat annotations support page.

Usage

Text Preprocessor

Wrapper for tokenized Wikipedia

brat annotations support

Examples

You can find them in the directory examples of our Toxine github repository.

License

Toxine is released under the BSD License. See the LICENSE file for more details.