A Survey on Comparable Entity Mining from Comparative Questions by Using Weakly Supervised and Markov-Logic Network

Abstract: Comparing two objects is a very typical part for human decision making process. However, this process is not always easy to know what to compare and what are the substitutes. To address this difficulty, we propose a novel way to automatically mine comparable entities from comparative questions that users posted online. To get high precision and high recall, we implemented a weakly-supervised bootstrapping method, to achieve comparable entity extraction and comparative question identification by leveraging a large online question archive. The experimental results show our proposed method achieves F1- measure of 82.5% in comparative question identification and 83.3% in comparable entity extraction. The comparative question identification and comparable entity extraction significantly outperform an existing state-of-the-art method.

Keywords: Automatically mine comparable entities, weakly-supervised bootstrapping method, F1- measure of 82.5%