Mining Comparators from Comparative Questions

Akash Saindanvise, Laxmi Venkatraman Tejas Shelke and Varun Varia

One of the essential parts of human life is to compare one thing with another in order to take proper decisions. But it is difficult to find what parameters to compare and what could be the alternatives for it. To solve this difficulty we present a novel way to mine comparable entities from comparative questions. To ensure that accuracy is maintained we develop a weakly supervised bootstrap method. Experimentation has shown that this method has achieved accuracy of about 82.5% in comparative question identification and 83.3% in extraction of comparable entities.The results are far better than the state of art system that exists.

Download PDF: