Perke is a Python keyphrase extraction package for Persian language. It provides an end-to-end keyphrase extraction pipeline in which each component can be easily modified or extended to develop new models.

• The easiest way to install is from PyPI:

  pip install perke

  Alternatively, you can install directly from GitHub:

  pip install git+https://github.com/alirezatheh/perke.git

• Perke also requires a trained POS tagger model. We use Hazm's POS tagger model. You can easily download latest Hazm's POS tagger using the following command:

  python -m perke download

  Alternatively, you can use another model with same tag names and structure, and put it in the resources directory.

Perke provides a standardized API for extracting keyphrases from a text. Start by typing the 4 lines below to use TextRank keyphrase extractor.

from perke.unsupervised.graph_based import TextRank

# 1. Create a TextRank extractor.
extractor = TextRank()

# 2. Load the text.
extractor.load_text(input='text or path/to/input_file')

# 3. Build the graph representation of the text and weight the
#    words. Keyphrase candidates are composed of the 33 percent
#    highest weighted words.
extractor.weight_candidates(top_t_percent=0.33)

# 4. Get the 10 highest weighted candidates as keyphrases.
keyphrases = extractor.get_n_best(n=10)

For more in depth examples see the examples directory.

Documentation and references are available at Read The Docs.

Perke currently, implements the following keyphrase extraction models:

• Unsupervised models
  • Graph-based models
    • TextRank: article by Mihalcea and Tarau, 2004
    • SingleRank: article by Wan and Xiao, 2008
    • TopicRank: article by Bougouin, Boudin and Daille, 2013
    • PositionRank: article by Florescu and Caragea, 2017
    • MultipartiteRank: article by Boudin, 2018

Perke is inspired by pke.