spacy_ke: Keyword Extraction with spaCy.

⏳ Installation

pip install spacy_ke

🚀 Quickstart

Usage as a spaCy pipeline component

import spacy

from spacy_ke import Yake

nlp = spacy.load("en_core_web_sm")
nlp.add_pipe(Yake(nlp))

doc = nlp(
    "Natural language processing (NLP) is a subfield of linguistics, computer science, and artificial intelligence "
    "concerned with the interactions between computers and human language, in particular how to program computers "
    "to process and analyze large amounts of natural language data. "
)

for keyword, score in doc._.extract_keywords(n=3):
    print(keyword, "-", score)

# computer science - 0.020279855002262884
# NLP - 0.035016746977200745
# Natural language processing - 0.04407186487965091

Customization

In the example below, we customize the yake algorithm as follows;

• Change the candidate selection to chunk (noun phrases). Notice that candidate_selection is a global config property for all keyword extractors, which can be set to either a callable (Doc -> Iterator[Candidate]), a string pointing to instance method (i.e chunk -> ._chunk_selection()), or a dict (i.e {"ngram": 3}).
• Set lemmatize=True for candidate weighting. Notice that this config property is unique to the Yake implementation.

nlp.add_pipe(Yake(nlp, candidate_selection="chunk", lemmatize=True))

Development

Set up pip & virtualenv

$ pipenv sync -d

Run unit test

$ pipenv run pytest

Run black (code formatter)

$ pipenv run black spacy_ke/ --config=pyproject.toml

References

[1] A Review of Keyphrase Extraction

@article{DBLP:journals/corr/abs-1905-05044,
  author    = {Eirini Papagiannopoulou and
               Grigorios Tsoumakas},
  title     = {A Review of Keyphrase Extraction},
  journal   = {CoRR},
  volume    = {abs/1905.05044},
  year      = {2019},
  url       = {http://arxiv.org/abs/1905.05044},
  archivePrefix = {arXiv},
  eprint    = {1905.05044},
  timestamp = {Tue, 28 May 2019 12:48:08 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-1905-05044.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

[2] pke: an open source python-based keyphrase extraction toolkit.

@InProceedings{boudin:2016:COLINGDEMO,
  author    = {Boudin, Florian},
  title     = {pke: an open source python-based keyphrase extraction toolkit},
  booktitle = {Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: System Demonstrations},
  month     = {December},
  year      = {2016},
  address   = {Osaka, Japan},
  pages     = {69--73},
  url       = {http://aclweb.org/anthology/C16-2015}
}