Automatic extraction of cause-effect relations in natural language text

The discovery of causal relations from text has been studied adopting various approaches based on rules or Machine Learning (ML) techniques. The approach proposed joins both rules and ML methods to combine the advantage of each one. In particular, our approach first identifies a set of plausible cause-effect pairs through a set of logical rules based on dependencies between words then it uses Bayesian inference to reduce the number of pairs produced by ambiguous patterns. The SemEval-2010 task 8 dataset challenge has been used to evaluate our model. The results demonstrate the ability of the rules for the relation extraction and the improvements made by the filtering process.