Rebuilding Knowledge Access through Natural Language Processing

Abstract: The vast size of the Internet has opened up immense possibilities for individuals to realize their potential. However, as the amount of information keeps growing exponentially, the tools we use to access knowledge are starting to fall behind. Tools like search engines and social media cannot precisely retrieve the desired knowledge, and often leading to information overload. With the breakthroughs in large-scale language models, I believe natural language processing (NLP) can revolutionize the way we access knowledge, making the process more efficient, precise and democratized.

In this talk, I will highlight my research on: 1) enhancing the encoding of language data to facilitate effective comprehension and retrieval of knowledge from text; 2) developing scalable and generalized reasoning machines that offer deep knowledge access; 3) and improving models’ generation through long-range knowledge reasoning. Finally, I will outline my future research goals that center on building more versatile knowledge machines that can eventually serve as an augmentation to human intelligence.

Bio: Wenhan Xiong is currently a senior research scientist at Meta AI working on Natural Language Processing (NLP). He earned his PhD in Computer Science from UC Santa Barbara in 2021 and his bachelor degree from the University of Science and Technology of China in 2016. With extensive experience in NLP, he has worked on a range of problems including Question Answering, Information Retrieval, Knowledge Graphs, Summarization, Information Extraction, and multi-modal tasks. He has published over 20 papers in prestigious AI conferences such as ICLR, ACL, EMNLP, NAACL, and ECCV. Additionally, he regularly serves as a reviewer for NLP and ML conferences, and has served as an area chair and session chair for EMNLP 2022 and ACL 2021.