Python implementation of TextRank, based on the Mihalcea 2004 paper.
Modifications to the original Mihalcea algorithm include:
- fixed bug; see Java impl, 2008
- use of lemmatization instead of stemming
- verbs included in the graph (but not in the resulting keyphrases)
- normalized keyphrase ranks used in summarization
The results produced by this implementation are intended more for use as feature vectors in machine learning, not as academic paper summaries.
Inspired by Williams 2016 talk on text summarization.
This code has dependencies on several other Python projects:
To install from PyPi:
pip install pytextrank
To install from this Git repo:
pip install -r requirements.txt
After installation you need to download a language model:
python -m nltk.downloader punkt python -m nltk.downloader wordnet python -m textblob.download_corpora python -m spacy.en.download all
Also, the runtime depends on a local file called stop.txt
which
contains a list of stopwords. You can override this in the
normalize_key_phrases() call.
See PyTextRank wiki
@htmartin @williamsmj @eugenep @mattkohl @vanita5 @HarshGrandeur @mnowotka