The search result page (SERP) of search engines has evolved fast in the past years. In addition to the traditional “ten blue links”, most commercial search engines provide direct answers to search queries which bear question intent. This feature of question answering (QA) greatly improves users experience by saving their efforts of clicking on the hyperlinks in SERP and searching for answers in web pages. With the wide adoption of voice search on mobile devices, more and more users tend to issue natural language questions to search engines and expect direct answers. We develop three question answering systems in Bing, namely, curated answering triggering, knowledge-based question answering, and passage-based question answering. In this talk, I will introduce the user experience of the three QA systems, illustrate several challenges, and present our approaches. In particular, I will also demonstrate how the recent DNN models and large-scale pre-trained models contribute to the QA tasks.
Microsoft Partner, Associate Director & Chief Scientist of Microsoft Software Technology Center Asia (STCA). Leads a team of 130+ scientists and engineers to build NLP applications and platform, mainly serving for Microsoft Bing and Cortana, and also contributing to Office, XiaoICE, and Microsoft Cognitive Services. Rich experience of research and engineering in Machine Learning, Data Mining, Natural Language Processing, and Bioinformatics. Ph.D. in Computer Science from the Statue University of New York at Buffalo. Tenure-track Assistant Professor in the Computer Science and Engineering School of Nanyang Technological University, Singapore (2005-2006), and Lead Researcher in Microsoft Research Asia (2007-2011). Published 30+ papers in top conferences and journals with 3000+ citations. Recipient of SIGKDD Best Application Paper Award in 2008, and Runner-up in 2004.