Causal Relation Classification using Convolutional Neural Networks and Grammar Tags

Identification of a cause-effect pair from nominal words is an important problem in knowledge reasoning of natural language texts. Most of the initial techniques for identifying cause effect relation are based on predefined linguistic and syntactic rules. Modern approaches uses Machine Learning techniques mainly Deep Neural Networks on top of sets of lingusitic rules and semantic knowledge to identify nominal word relations in a text. In this paper, a novel algorithm has been proposed using convolutional neural networks and grammar tags. The proposed approach has been evaluated using SemEval-2010 Task 8 dataset and shown to perform better than the other state-of-the-art algorithms.