Jing Liu（刘璟）

Principal Architect
Natural Language Processing Department at Baidu Inc.


I am a principal architect and a tech lead of deep question answering team at Baidu NLP since December 2017. Before that, I was a researcher at Microsoft Research Asia (MSRA) from September 2014 to December 2017. I obtained Ph.D. degree in computer science from Harbin Institute of Technology (HIT) under the supervision of Prof. Hsiao-Wuen Hon (MSRA), Prof. Ting Liu (HIT) and Dr. Chin-Yew Lin (MSRA) in September 2014. My research interests include question answering, information extraction and social computing.

Please contact me via legendarydan (at) gmail (dot) com
My Sina Weibo (in Chinese) and Twitter (in English)

News

  • Please send me emails with your resume (for internships or FTE positions) if you are interested in working with us on question answering and machine reading comprehension. Experiences with machine (incl. but not limited to deep) learning for NLP are preferred.
  • Oct 2020: We proposed RocketQA [paper], an optimized training approach to dense passage retrieval for open-domain question answering. RocketQA achieved the 1st rank at the leaderboard of MSMARCO Passage Ranking Task. It was featured in zh-cn and en-us.
  • Aug 2020: Baidu, CCF (China Computer Federation) and CIPSC (Chinese Information Processing Society of China) jointly lunched the project of LUGE（千言）[portal], that is an open-source project of Chinese NLP benchmarks. Our aim is to promote the advancement of Chinese NLP technologies by the new benchmarks. Specifically, LUGE tries to evaluate models beyond just accuracy, in terms of robustness, generalization, multi-task capabilities etc., and cover rich types of tasks, including the tasks of language understanding, language generation and multimodality. Currently, we have collected more than 20 NLP datasets for 7 tasks from the great contributors of 11 organizations. LUGE was featured in videos (zh-cn, en-us) and articles (zh-cn). If you are interested in LUGE, pls. contact me.
  • Apr 2020: We released a Chinese dataset namely DuReaderrobust [paper][data & code] towards evaluating the robustness of machine reading comprehension models. We hosted a shared task of DuReaderrobust [leaderboard] at 2020 Language and Intelligence Challenge, and there were more than 1,500 teams and more than 4,600 submissions in the shared task. The shared task was featured in zh-cn.
  • Nov 2019: Our proposed machine reading comprehension system D-NET [paper][code] was ranked at top 1 in the MRQA 2019 Shared Task, that tests if MRC systems can generalize beyond the datasets on which they were trained. D-NET was featured in zh-cn (1, 2) and en-us.

Professional Activities

  • Area Chair: ACL 2021 (Question Answering)
  • Session Chair: AACL 2020 (Question Answering)
  • Program commitee/reviewer, ACL, EMNLP, NAACL, EACL, AACL, SIGIR, KDD, WSDM, WWW, CIKM, ICWSM, ACM Transactions on the Web (TWEB), ACM Transactions on Intelligent Systems and Technology (TIST), ACM Transactions on Information Systems (TOIS), Frontiers of Computer Science (FCS)

Papers [Google Scholar]

Working Experience

  • Principal Architect, Baidu NLP, Dec. 2017 - present
  • Researcher, Microsoft Research Asia, Sep. 2014 - Dec. 2017
  • Intern, Microsoft Research Asia, Jul. 2009 - Sep. 2014

Educations

  • PhD, Computer Science, Harbin Institute of Technology, Sep. 2009 - Sep. 2014
  • M.Sc, Computer Science, Harbin Institute of Technology, Sep. 2007 - Jul. 2009
  • B.Sc, Computer Science, Xidian University, Sep. 2003 - Jul. 2007

Last updated Nov 20 2020 (This template was originally designed by Mu Li.)