Dr. Lu Chen joined the Department of Computer Science and Engineering, Shanghai Jiao Tong University (SJTU) as an assistant research professor in October 2020. He obtained his Ph.D. degree from Shanghai Jiao Tong University in June 2020 and his doctoral thesis was awarded as SJTU Outstanding Doctoral Dissertation. His research interests include dialogue systems, question answering, and natural language processing. The goal of his research is to build evolvable and universal conversational agents, which can converse with humans among many domains and improve their performance with various signals. He has authored/co-authored more than 30 journal articles (e.g. IEEE/ACM transactions) and peer-reviewed conference papers (e.g. ACL, EMNLP, AAAI, COLING), one of them was selected as COLING2018 Area Chair Favorites.
Recently, conversational natural language interface to databases (NLIDB) has attracted lots of interest in both academic and industrial communities. It provides a convenient method for users to interact with databases. The core technology behind NLIDB is Text-to-SQL, which aims to convert a natural language question into a SQL query, given the corresponding database schema. One daunting challenge for Text-to-SQL is how to jointly encode the question words and database schema items (including tables and columns), as well as various relations among these heterogeneous inputs. In this talk, we will try to answer this question in two respects: First, we will introduce our proposed ShadowGNN, which processes database schemas at abstract and semantic levels and aims to improve the generalization capability for rare and unseen database schemas. Second, we will introduce the line graph enhanced Text-to-SQL model, which has the ability to model local and non-local relations in heterogeneous inputs.