Skip to main content

textcrafts: Summary, keyphrase and relation extraction with dependecy graphs

Project description


Python-based summary, keyphrase and relation extractor from text documents using dependency graphs.


Project Description

** The system uses dependency links for building Text Graphs, that with help of a centrality algorithm like PageRank, extract relevant keyphrases, summaries and relations from text documents. Developed with Python 3, on OS X, but portable to Linux.**


  • python 3.7 or newer, pip3, java 9.x or newer. Also, having git installed is recommended for easy updates
  • pip3 install nltk
  • also, run in python3 something like
import nltk'wordnet')'words')'stopwords')
  • or, if that fails on a Mac, use run python3 to collect the desired nltk resource files.
  • pip3 install networkx
  • pip3 install requests
  • pip3 install graphviz, also ensure .gv files can be viewed
  • pip3 install stanfordnlp parser
  • Note that stanfordnlp requires torch binaries which are easier to instal with ````anaconda```.

Tested with the above on a Mac, with macOS Mojave and Catalina and on Ubuntu Linux 18.x.

Running it:

in a shell window, run

in another shell window, start with

python3 -i

and then interactively, at the ">>>" prompt, try

>>> test1()
>>> test2()
>>> ...
>>> test9()
>>> test12()
>>> test0()

see how to activate other outputs in file

text file inputs (including the US Constitution const.txt) are in the folder


Handling PDF documents

The easiest way to do this is to install pdftotext, which is part of Poppler tools.

If pdftotext is installed, you can place a file like textrank.pdf already in subdirectory pdfs/ and try something similar to:

Change setting in file to use the system with other global parameter settings.

Alternative NLP toolkit

Optionally, you can activate the alternative Stanford CoreNLP toolkit as follows:

  • install Stanford CoreNLP and unzip in a derictory of your choice (ag., the local directory)
  • edit if needed with the location of the parser directory
  • override the params class and set corenlp=True

Note however that the Stanford CoreNLP is GPL-licensed, which can place restrictions on proprietary software activating this option.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

textcrafts-0.1.2.tar.gz (13.4 kB view hashes)

Uploaded source

Built Distribution

textcrafts-0.1.2-py3-none-any.whl (17.9 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page