报告题目: Bridging the Gap between AI and SE
报告人：夏鑫 助理教授（讲师） 澳大利亚蒙纳士大学
邀请人：郑炜 副教授 软件学院
Xin Xia is a lecturer with tenure (a.k.a. Assistant Professor) at the Faculty of Information Technology, Monash University, Australia. Prior to joining Monash University, he was a post-doctoral research fellow working with Prof. Gail Murphy in the software practices lab at the University of British Columbia. Dr. Xia received his Ph.D. from Zhejiang University in 2014. His current research focuses on analyzing the rich data in software repositories to uncover actionable information that can improve the productivity of developers and testers.
Dr. Xia has published 135 papers with 105 different co-authors, including 23 CCF A Conference/Journal Papers (e.g., 12 TSE, 1 TOSEM, 3 ICSE, and 5 ASE papers). His work has received several best/distinguished paper awards, e.g., ACM SIGSOFT Distinguished Paper Award at ASE 2018 and ICPC 2018, FOSS Impact Paper Award at MSR 2018. Dr. Xia is among the top two rising software engineering (SE) researchers in the world and is in the top 5 most active early career SE researchers worldwide according to an analysis of the publications of SE scholars during the last decade . Dr. Xia serves on the SANER steering committee. His service includes serving as an organizer of SANER 2019, ICECCS 2019, and ASE 2016, and as the program committee of Software Engineering conferences and journals, e.g., ICSE 2019-2020, ASE 2019, ICSME 2017-2019, MSR 2017-2019, TSE, TOSEM, and EMSE. Dr. Xia won the distinguished reviewer award from the EMSE journal in 2019.
Today, data miners often apply or extend AI techniques to solve problems across many domains (e.g., social media, health informatics, and software systems); while domain experts leverage their own domain knowledge to solve their own problems. Data miners often apply their automated techniques to solve a wide range of problems across different domains with limited knowledge of the domain; while domain experts often have limited knowledge of automated techniques when solving their domain-specific problems.
My research tries to bridge the gap between both types of experts (i.e., Data miners and Domain Experts). In this talk, I will focus on the software engineering domain and I will give an overview of several challenges facing data miner and domain experts as they make use of automated techniques, in particular:(1) strong performance of techniques is not sufficient, instead a deeper understanding of the domain is essential; (2) results should be presented in a domain-centric context; (3) an easy approach might perform better than a complex approach. I will present examples from my research to explain what these challenges are, why do they appear, and my efforts to avoid them.