README.rst

Nerdy provides different NER classifiers as a service in a common format, NIF. Users can extract named entities from text they provide or use annotated documents to evaluate and compare the classifiers.

Nerdy uses Senpy, a software to create sentiment analysis web services easily developed at GSI UPM.

Available classifiers

• Stanford classifier : Spanish and English language.
• Polyglot classifier : Spanish and English language.
• Citius classifier: For Spanish and English language.
• Ritter classifier: Only for english text.

Installation

First, you need to install senpy. You could do this through PIP.

pip install --user senpy

Then, you clone this repository and run a senpy server with nerdy plugin.

git clone http://github.com/gsi-upm/nerdy
cd nerdy
python2 -m senpy -f .

The service is now available at port 5000.

Known issues

• This software is developed in Python 2.7. So, the release of Senpy installed in your host must be python 2 release. (don't use pip3)
• To use Standford classifier with Nerdy, Oracle Java 8 is neccesary.
• To use citius NER, please visit http://proxectos.citius.usc.es/hpcpln/index.php/en/software-tools. Move Citius_tools folder into classifiers folder.
• NER run files, datasets and classifiers needed to use with this tool are not included in this repo. To use them, please visit official sites as https://nlp.stanford.edu/software/CRF-NER.shtml
• It is possible you need to install some extra depedencies for using polyglot. You can do this on Linux with:

sudo apt-get install libicu-dev python-numpy