HBCP Semantic Prediction

The repository for the HBCP semantic-enhanced prediction system.

We aim to develop an interpretable system for prediction of outcomes from behaviour change interventions using the Behaviour Change Intervention Ontology and a corpus of annotated literature.

How to run the rule prediction

1. Generate the dataset: python dataprocessing/dataprocessing.py This should create a pickled version of the dataset in data/hbcp_gen.pkl

2. Run the model: python rulenn/rule_nn.py data/hbcp_gen.pkl

The rules that have been trained should then be available under results/rules.txt. Rules are separated by semicolons. The last element of each rule is the rule weight.

Data cleaning

The mechanisms used for data cleaning are documented in docs/cleaner.md