News

• I give a tutorial about ``Deep Chit-Chat: Deep Learning for ChatBots" with Rui Yan on EMNLP 2018. Please find the slides in Talks.
• Our paper "Multi-Representation Fusion Network for Multi-turn Response Selection in Retrieval-based Chatbots" is accepted by WSDM'19. In this paper, we study how to leverage multiple types of representations for context-response matching. We get 2.1% improvement on R10@1 over the state-of-the-art model DAM on the Ubuntu data, and 2.1% improvement on P@1 over DAM on the Douban data.

About Me

I am a principal applied scientist lead in Microsoft Xiaoice Team since January, 2018. Before that, I was a lead researcher in Microsoft Research Asia (MSRA). I joined MSRA in 2012 and had been a member of Natural Language Computing (NLC) group until December 2017. Before I became an employee in MSRA, I was a joint student of MSRA and Peking University. My advisor is Professor Hang Li.

I obtained a B.S. in Applied Mathematics from Peking University in 2007 and earned my Ph.D. in Applied Mathematics from Peking University in 2012.

My research interests include machine learning, natural language processing, and information retrieval.

My current research focus is building conversational engines for chatbots with machine learning and NLP techniques. I have been working on single-turn conversation, multi-turn conversation, and topical deep chat since 2014. I am a key technology contributor to core chat engines of Microsoft XiaoIce (微软小冰) v2, v3, v4, v5 and v6 and Microsoft Rinna (りんな). My significant achievements include the launching of generation based chat engines in XiaoIce and Rinna and the recent Empathy Model in Rinna. Particularly, my work serves Microsoft's chatbot in Indonesia with full dialogue generation technologies. The chatbot now has more than 1.5 million users on LINE Indonesia.