Xinwei Geng's Homepage

Xinwei Geng

I am a Ph.D. student in Research Center for Social Computing and Information Retrieval at Harbin Institute of Technology (HIT, China) since 2015. My advisor is Prof. Bing Qin. My research interests focus on Natural Language Processing, Text Generation, Neural Machine Translation and Reinforcement Learning. Specially, I'm interested in improving neural machine translation with reinforcement learning. I obtained B.S. degree in computer science from HIT in 2014.

Email contact: xwgeng (at) ir (dot) hit (dot) edu (dot) cn
Find me in: Github , Weibo

August 26, 2021. One paper has been accepted by EMNLP 2021.
May 1, 2021. One paper has been accepted by IJCAI 2021.
April 4, 2020. One paper has been accepted by ACL 2020.
June 25, 2018. Research Intern at Tencent AI Lab, Supervised by Dr. Zhaopeng Tu and Dr. Longyue Wang

Ph.D candidate, Harbin Institute of Technology, September 2015 - present
- Major: Computer Science
- Advisor: Bing Qin, Professor in Harbin Institute of Technology
M.S., Harbin Institute of Technology, September 2014 - July 2015
- Major: Computer Science
- Advisor: Bing Qin, Professor in Harbin Institute of Technology
B.S., Harbin Institute of Technology, September 2010 - July 2014
- Major: Computer Science
- Score: 90.61(Top 6%, Rank 12/203)

Learning to Rewrite for Non-Autoregressive Neural Machine Translation
Xinwei Geng, Xiaocheng Feng, Bing Qin
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP 2021, full paper)

Dialogue Discourse-Aware Graph Model and Data Augmentation for Meeting Summarization [code] [arXiv]
Xiachong Feng, Xiaocheng Feng, Bing Qin, Xinwei Geng
Proceeding of The 30th International Joint Conference on Artificial Intelligence (IJCAI 2021)

How Does Selective Mechanism Improve Self-Attention Networks? [pdf] [bib] [poster] [slides]
Xinwei Geng, Longyue Wang, Xing Wang, Bing Qin, Ting Liu, Zhaopeng Tu
Proceedings of The 58th Annual Meeting of the Association for Computational Linguistics (ACL 2020, full paper)

Learning to Refine Source Representations for Neural Machine Translation. [arXiv]
Xinwei Geng, Longyue Wang, Xing Wang, Bing Qin, Ting Liu, Zhaopeng Tu

Adaptive Multi-pass decoder for Neural Machine Translation. [pdf] [bib] [poster] [slides]
Xinwei Geng, Xiaocheng Feng, Bing Qin, Ting Liu
Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP 2018, full paper)

A Planning based Framework for Essay Generation. [arXiv]
Bing Qin, Duyu Tang, Xinwei Geng, Dandan Ning, Jiahao Liu, Ting Liu

December 2019 - Present. Peng Cheng Laboratory's Machine Translation System
- The target of this project is to build the machine translation system for the Belt and Road Initiative
- Develop the Peng Cheng Laboratory's Machine Translation System, which can complete the translation
  between other language and Chinese.
June 2015 - July 2018. National High Technology Development 863 Program of China
- The target of this project is to solve the questions and generate the corresponding answers
  with natural language processing automatically for Gaokao
- Develop the Topic-to-Essay Generation System, which integrates extractive and abstractive approches together
September 2014 - July 2015. Sogou's University Student Innovation Program
- Extract synonyms given the name entities from online baike corpus
- Obtain 200,000 synonym pairs related to 120,000 name entities from 60,000 baike pages
September 2013 - July 2014. The BigCilin System
- BigCilin is a large-scale Chinese open domain hyponymy-based knowledge graph,
  which dynamically constructs its hierarchical architecture of classes based on search engine, online baike and raw texts.
- Develop the module for synonyms extraction from large-scale raw text
- BigCilin has been purchased by many companies such as Tencent, Qihoo 360, IFLYTEK

Conference Reviewer/Secondary Reviewer: ACL 2021, AAAI (2019, 2021), NAACL 2019, CCL (2019, 2020)

June 2018 - July 2019 Full-time intern at Tencent AI Lab, Supervised by Dr. Zhaopeng Tu and Dr. Longyue Wang