Can Taxonomy Help? Improving Semantic Question Matching using Question\n Taxonomy

In this paper, we propose a hybrid technique for semantic question matching.\nIt uses our proposed two-layered taxonomy for English questions by augmenting\nstate-of-the-art deep learning models with question classes obtained from a\ndeep learning based question classifier. Experiments performed on three\nopen-domain datasets demonstrate the effectiveness of our proposed approach. We\nachieve state-of-the-art results on partial ordering question ranking (POQR)\nbenchmark dataset. Our empirical analysis shows that coupling standard\ndistributional features (provided by the question encoder) with knowledge from\ntaxonomy is more effective than either deep learning (DL) or taxonomy-based\nknowledge alone.\n