This project contains training data and procedure of NLP model for relation extraction, web service that uses model for news processing and Chrome extension that shows relations to users.
For traning procedure see "Model training" Jupyter notebook in "Model training" folder. Training data files in same folder contain preprocessed and annotated training sentences from Reuters and Brown datasets.
See semantic.py in "Web service" folder. Evaluation results with used texts are in "Web service/evaluation" folder.
Entrypoint is relation_extractor_web.py in "Web service" folder. Service requires running MySQL instance. Check database_mysql.py for details.
Source is in "Chrome extension" folder. It could be easily installed by enabling "Developer mode" at chrome://extensions/ and selecting "Load unpacked".