Python library for concept graph extraction from text, operation, and linearization. An integrated web service is provided.

This library is still a work in progress, but it has shown to be already useful for a number of applications, for example extractive text summarization.

Using uv:

1. uv sync
2. If you want to install some extras, run uv sync --extras "web lexical modules" with the extras that you need (see pyproject.toml).
3. uv run setup

To run any grafeno script with poetry, use uv run before the name of the script and its arguments.

The documentation is a work in progress, so it is a bit patchy, but go ahead and read it in ReadTheDocs.

See the notebooks in the examples directory for how to use grafeno in different applications.

Run the server.py script to get a json web service which exposes most of the pipeline functionality.

Use -h to get the list of options available.

A test script is provided in test.py that can run a pipeline to test the library. It can serve as the entry point to the library operation, or as an example of how to use it from python.

Use -h to get the list of options available.

• python >= 3.4
  • Python packages for use of the library are listed in requirements.txt. We recommend using conda to install grafeno and its dependencies in a virtual environment.
• A dependency parser. For now, the following are supported:
  • spaCy (recommended)
  • freeling
• If using the simplenlg linearizer, a java executable will have to be available.

You may also need some NLTK data, for example 'wordnet' and 'wordnet_ic'. They can be downloaded in python with:

import nltk
nltk.download(['wordnet', 'wordnet_ic'])

• Antonio F. G. Sevilla [email protected]
• Alberto Díaz [email protected]

The continued development of this library has been possible thanks to a number of different research and development projects, listed below.

• A collaboration with MedWhat, a company that develops virtual medical assistant bots and other medical artificial intelligence solutions.
• This research is funded by the Spanish Ministry of Economy and Competitiveness and the European Regional Development Fund (TIN2015-66655-R (MINECO/FEDER)).
• This work is funded by ConCreTe. The project ConCreTe acknowledges the financial support of the Future and Emerging Technologies (FET) programme within the Seventh Framework Programme for Research of the European Commission, under FET grant number 611733.