中央研究院 資訊科學研究所

活動訊息

友善列印

列印可使用瀏覽器提供的(Ctrl+P)功能

New Progress of Knowledge-Based Question-Answering

:::

New Progress of Knowledge-Based Question-Answering

  • 講者周明 博士 (微軟亞洲研究院)
    邀請人:蘇克毅
  • 時間2016-10-04 (Tue.) 10:00 ~ 12:00
  • 地點資訊所新館106演講廳
摘要

Knowledge-Based Question Answering (KB-QA)is a significant human-computer interaction technology for search engine, database, speech assistant, chat-bot and has broad applications in mobile internet, e-commerce, customer support and IoT. In recent years, driven by the strong needs of search engine and mobile internet and supported by the availability of large scale knowledge bases, KB-QA has made considerable progress and various novel methodologies have emerged. However, though we have seen plenty of exciting progress of KB-QA, there are still many challenging problems unsolved. The purpose of this talk is to carefully study the recent progress of KB-QA, analyze its challenges and accordingly propose new approaches in order to further boost it to the next level. In this presentation, I will start with an introduction to KB-QA including its definition, the knowledge bases readily available for KB-QA and an brief summary of its technical development in last 30 years. Then, I will elaborate two major technological approaches of KB-QA including semantic parser motivated QA (SP-QA) and IR motivated QA (IR-QA). The effective application of Deep Learning technologies demonstrated in recent work of KB-QA will also be described. To understand the advantage and limitations of the existing approaches of KB-QA, I made a solid comparison of various systems based on their performance on a public evaluation dataset and propose my suggestions on addressing the observed problems in the future. In the final part of this talk, I will introduce the work on KB-QA conducted at Microsoft Research Asia including the LIGHT QA system and the TRINITY graph engine which is used for the efficient storage and access of large scale KB.

 

BIO

Dr. Ming Zhou is a principal researcher and manager of Natural Language Computing Group in Microsoft Research Asia.  He is the chair of Chinese Information Technology Committee of Chinese Computer Federation and executive member of Chinese Information Processing Society.

He designed the CEMT-I machine translation system in 1989, the first experiment of Chinese-English machine translation in Mainland China. He designed the famous Chinese-Japanese machine translation software product J-Beijing in Japan which was deployed in J-Server, the popular translation service in Japan that was granted Makoto Nagao Award by Japan Machine Translation Association in 2008. He is the leader of the famous AI gaming of Chinese Couplets/Poetry Generation and Riddles(http://duilian.msra.cn), and the English Assistance Search Engine, Engkoo, which won the Wall Street Journal’s 2010 Asian Innovation Readers’Choice Award and was shipped in Bing in 2011 as Bing Dictionary(http://cn.bing.com/dict/), and Engkoo cloud IME which was shipped as Bing IME in 2012. Recently, his group has closely worked with MS product teams and shipped famous chat-bot products in China(Xiaoice), Japan(Rinna) and US(Tay).

Dr. Zhou received his B.S. degree in computer engineering from Chongqing University in 1985, and his M.S. degree and Ph.D. in computer science from Harbin Institute of Technology in 1988 and 1991. He did post-doctoral work at Tsinghua University from 1991 to 1993, then he became an associate professor. During 1996-1999, during his sabbatical leave, he worked for Kodensha Ltd. Co. in Japan as the leader of the Chinese-Japanese machine translation project. He joined the natural language group at Microsoft Research China (now Microsoft Research Asia) in Sept. 1999.