The Report Data：18 October

Time	Speaker	Topic
19:00—19:45	Wanxiang Che	How to Make a Wonderful Academic Report
19:45—20:30	Kang Liu, Xianpei Han	10 Methods of Doing Failed Research
20:30—20:50		Tea Break
20:50—21:10	Jingjing Xu	Robustness Analysis of Machine Learning Based on Antagonism Training - Craggy and Thinking in Research
21:10—21:30	Xu Han	Knowledge Representation and Acquisition -- Large System and Small Cooperation in Doctoral Career
21:30—22:00		Guest Forum - Open Discussion Between Guest and Audience

Report Details：

Time：   19:00—19:45
Report Title：How to Make a Wonderful Academic Report
Speaker：    Wanxiang Che
Report Summary：With the gradual improvement of China's scientific research level, more and more students have the opportunity to publish papers in high-level academic conferences at home and abroad.In order to make the paper published smoothly, students often attach great importance to the training of writing ability, and often neglect the report link of the paper.In fact, publishing a paper is not the last step of scientific research. In order to further expand the influence of the results and let more scholars understand our work, we must attach importance to every opportunity to do academic reports.Therefore, we will summarize some experience of how to do a good academic report, including the purpose of the report, the psychology of the audience, the content presented in the report, the format of the slides and some points to be noted in the process of the report, and finally hope to help you to make a wonderful academic report.

Wanxiang Che :Ph.D., school of computer science, Harbin Institute of Technology, doctoral supervisor, visiting scholar at Stanford university, and co-supervisor, professor Christopher Manning.Currently, he is a member of the computational linguistics professional committee of the Chinese information society of China, deputy director of the youth working committee, and a senior member of the Chinese computer society of China. He once served as the chairman of Harbin YOCSEF (2016-2017) in ACL, EMNLP, AAAI,IJCAI high level at home and abroad such as journal published more than 50 papers and conference, AAAI 2013 articles won best dissertation nomination, paper has cited more than 2100 times (by Google Scholar), 2 H - the index values of 26 published materials, such as translation 2 bear the national natural science fund of 973 a number of scientific research project is responsible for research and development of language technology platform has been more than 600 units (LTP) sharing, online language provided by the cloudThe service has more than 10,000 users, and has been authorized to baidu, tencent, huawei and other companies for use in 2018. It has won the first place in the international evaluation of CoNLL multilingual syntactic analysis from 2015 to 2016, and won the Google Focused Research Award(Google Focused Research Award) for two consecutive years.In 2016, won the first prize of science and technology progress of heilongjiang province (ranking 2nd);In 2012, won the second prize of heilongjiang provincial technological invention award (ranking the 2nd);In 2010, he won the first prize (ranked 2nd) of the first hanwang youth innovation award (individual) and other awards. In 2017, his MOOC course "high-level language programming (Python)" won the national high-quality online open course.

Time：   19:45—20:30
Report Title：10 Methods of Doing Failed Research
Speaker：    Kang Liu, Xianpei Han
Report Summary：Successful research are similar, failure research each have each successful research methods allow you to look up at the starry sky, failure research methods make you when looking at the stars to avoid falling into the pool if successful scientific research is to learn, so is case make you more good papers (Recall), but we also need a lot of negative example let you use the less contribute more good papers (Precision), little detours, our optimization goal is to find the optimal F value of this report is based on the two speakers in 30 years (sum) had seen in scientific research experienceWhat we've heard about failed research methods, including the philosophical mindset of failed research ideas, methods, techniques and tools we try to include a lot of information, but at the same time keep it interesting and of course, at the end of describing 10(and maybe more than) failed research methods, we also give the opposite of failed research how to make a successful research.

Kang Liu :Ph.D., the vice Researcher of State Key Laboratory of pattern recognition of Chinese Academy of Sciences, visiting professor of Xidian University. and his research field includes information extraction, web mining and question answering system, etc., but also involves the basis of pattern recognition and machine learning research. In the field of natural language processing such as knowledge engineering international important conferences and journals published more than 90 papers (such as TKDE ACL IJCAI EMNLP COLING CIKM, etc.), obtain the KDD CUP 2011 track 1 second, global COLING, The first CCF best paper Award, 2014 - Tencent Rhino Bird Fund Excellence Award, 2015, 2016, Google Focused Research Award, China Chinese information society in 2014, the author - hanwang youth innovation of the Chinese information processing science and technology prize first prize, China Chinese information society in 2018, the author first prize (second) of the Chinese information processing science and technology prize awards, and hold a concurrent post as the director of China Chinese Information Society Youth Work Committee and the Secretary general of Language and Knowledge Computing Committee.

Xianpei Han :Ph.D.,the researcher of State Key Laboratory of Computer Science,the main research direction for information extraction, knowledge map, semantic parsing and intelligent question answering system in the ACL, SIGIR who lead, AAAI, EMNLP published more than 40 papers, and other important international conference, Xianpei Han is director of Chinese Information Processing Society of China, Language and Knowledge Computing Professional Committee, deputy director of the green meeting member, Chinese academy of sciences, 2016, included in the lift for young scientists of China association for science and technology plan, hanwang youth innovation of China Chinese information society.

Time：   20:50—21:10
Report Title：Robustness Analysis of Machine Learning Based on Antagonism Training - Craggy and Thinking in Research
Speaker：    Jingjing Xu
Report Summary：With the development of deep learning, the neural network model is constantly refreshing the best effect in various fields. However, the robustness of the current neural network to noise is poor. Therefore, how to improve the robustness of machine learning is a problem that must be solved to transfer the model from the laboratory environment to the real environment.This report shows how to automatically generate samples that cannot be classified correctly by the model through the method of confrontation training, so as to improve the robustness of the model. In addition, I will take this study as an example to introduce my setbacks and thoughts during the scientific research

Jingjing Xu :He is a 15-year doctoral candidate in the school of information science and technology, Peking University. His supervisor is Sun Xu, and his research direction is natural language processing.During his doctoral study, he published 7 research papers in ACL, EMNLP, NAACL and other top natural language processing conferences and journals as the first author, and many research papers as the co-author. he has won honors and awards such as merit student, merit student model, Canon scholarship, and future star of Microsoft Asia research institute.In addition to solving field research problems, he also participated in a number of actual landing projects, including the development of multi-domain word segmentation tool pkuseg, participated in the JDDC global task-oriented multi-round man-machine dialogue challenge and won the champion of group automatic evaluation and the second place of manual evaluation.

Time：   21:10—21:30
Report Topic：Knowledge Representation and Acquisition -- Large System and Small Cooperation in Doctoral Career
Speaker：    Xu Han
Report Summary：Introducing structured knowledge is one of the important methods to assist natural language processing tasks.How to accurately obtain structured information from free texts and how to effectively represent knowledge has gained wide attention in recent years.In this report, the speakers will comb the development of the knowledge representation and acquisition, share the latest progress in the field of related, Chen will be with him in the knowledge representation and relationship extraction on some representative work as an example, to deeply explore the specific problems in the research of analysis, combined with the speakers personal working experience, discuss how to carry out systematic research and academic cooperation and other issues, share its some experience in the process to solve the problem.

Xu Han :He is a 17-year doctoral candidate in the department of computer science, Tsinghua university. He is from the natural language processing group of Tsinghua University, under the guidance of associate professor Zhiyuan Liu. His main research direction is natural language processing and information extraction. He has published many papers related to ACL, EMNLP, NAACL, COLING and AAAI, and maintained many open source projects on Github.