ÀÏÍõÂÛ̳

Speaker: , Ph.D. candidate, Texas A&M University

 

Title: Addressing Information Distortion via Knowledge-aware Language Modeling 

  

In the research of Natural Language Processing (NLP), language and information are deeply intertwined. However, human language can be subject to various forms of information distortion, ranging from intentional misinformation to unintentional bias. Information distortion is pervasive across human communications and poses significant threats to society. Beyond simple binary of true or false, information distortion often involves nuanced layers of slanted framing or flawed reasoning to strategically manipulate readers without relying on outright falsehoods. 

In this talk, I will present my research on innovating knowledge-aware language modeling for detecting and mitigating information distortion. I will share my vision for equipping large language models (LLMs) with structured knowledge, enabling them to capture the intricate connections between different pieces of information. I will introduce various types of structured knowledge, and delve into the algorithms for constructing, representing, and integrating these structured knowledge into LLMs. Finally, I will discuss how knowledge-aware language modeling can be adapted to a broader range of NLP tasks and multi-modality learning scenarios that demand complex knowledge-based reasoning. 

 

Bio: Yuanyuan Lei is a final-year PhD candidate in the Department of Computer Science at Texas A&M University. Her research centers around Natural Language Processing and Large Language Models. In particular, her research innovates knowledge-aware language modeling for detecting and mitigating information distortion in human language. Her research has been published in top-tier Natural Language Processing venues, including EMNLP, ACL, NAACL. She was awarded with Future Research Leaders in Artificial Intelligence by the University of Michigan, and Rising Star in Data Science by the University of Chicago.