Real, Live, and Concise: Answering Open-Domain Questions with Word Embedding and Summarization

Resorting to community question answering (CQA) websites for finding answers has gained momentum in the recent years with the explosive rate at which social media has been proliferating. With many questions left unanswered on those websites, automatic question answering (QA) systems have seen light. A main objective of those systems is to harness the plethora of existing answered questions; hence transforming the problem to finding good answers to newlyposed questions from similar previously-answered ones or composing a new concise one from those potential answers. In this paper, we describe the real-time Question Answering system we have developed to participate in TREC 2016 LiveQA track. Our QA system is composed of three phases: answer retrieval from three different Web sources (Yahoo! Answers, Google Search, and Bing Search), answer ranking using learning to rank models, and summarization of top ranked answers. Official track results of our three submitted runs show that our runs significantly outperformed the average scores of all participated runs across the entire spectrum of official evaluation measures deployed by the track organizers this year.