Talk: Learning to Build Conversational Natural Language Interfaces

Abstract: Natural language is a fundamental form of information and communication and is becoming the next frontier in computer interfaces. Natural Language Interfaces (NLIs) connect the data and the user, significantly promoting the possibility and efficiency of information access for many users besides data experts. All consumer-facing software will one day have a dialogue interface, the next vital leap in the evolution of search engines. Such intelligent dialogue systems should be able to understand the meaning of language grounded in various contexts and generate effective language responses in different formats for information requests and human-computer communication.

In this talk, I will cover three key developments that present opportunities and challenges for the development of deep learning technologies for conversational natural language interfaces. First, I will discuss the design and curation of large datasets to drive advancements towards neural-based conversational NLIs. Second, I will describe the development of scalable algorithms to parse complex and sequential questions to formal programs (e.g. mapping questions to SQL queries that can execute against databases). Third, I will discuss the general advances of language model pre-training methods to understand the meaning of language grounded in various contexts (e.g. databases and knowledge graphs). Finally, I will conclude my talk by proposing future directions towards human-centered, universal, and trustworthy conversational NLIs.

Bio: Tao Yu is a fourth-year Ph.D. candidate in Computer Science at Yale University. His research aims to build conversational natural language interfaces (NLIs) that can help humans explore and reason over data in any application (e.g., relational databases and mobile apps) in a robust and trusted manner. Tao’s work has been published at top-tier conferences in NLP and Machine Learning (ACL, EMNLP, NAACL, and ICLR). Tao introduced and organized multiple popular shared tasks for building conversational NLIs, which have attracted more than 100 submissions from top research labs and which have become the standard evaluation benchmarks in the field. He designed and developed language models that achieve new state-of-the-art results for seven representative tasks on semantic parsing, dialogue, and question answering. He has worked closely with and mentored over 15 students and collaborated with about 20 researchers from Salesforce Research, Microsoft Research, Columbia University, UC Berkeley, the University of Michigan, and Cornell University. He has been on the program committee for about ten NLP and Machine Learning conferences and workshops, including one of the main organizers of the Workshop of Interactive and Executable Semantic Parsing at EMNLP 2020. For more details, see Tao’s website: https://taoyds.github.io.