Long Narrative Understanding in the Era of Large Language Models
Large Language Models (LLMs) have demonstrated impressive proficiency in comprehending human language. However, they face challenges in processing long and complex text such as ESG reports and novels, and engaging in prolonged, multi-topic conversations. In this seminar, I will present the latest research from my group, which delves into the complexities of comprehending lengthy narratives. In particular, I will present our innovative and scalable Table-of-Contents extraction framework for ESG reports, novel strategies for tuning LLMs to utilise dynamic memos to ensure coherent long-range conversation, and a unique system enabling users to immerse themselves as fictional characters and interact with other characters within novels.

I will conclude my talk with a discussion of challenges associated with understanding extended narratives and outline potential directions for future research in this ever-evolving field.

Speaker
Yulan He is a Professor at the Department of Informatics. Her research centres around the exploration of statistical models in representing uncertainty and the benefit they bring over earlier work in a wide range of application areas, particularly the integration of machine learning and natural language processing for text understanding. She currently holds a five-year Turing AI Fellowship, funded by the UK Research and Innovation (UKRI).