Professor Joemon Jose‘s talk
Date and time: Wednesday Sep 27, 2023 15:00-16:00
Revisiting Legal Citation Prediction in the Era of Emerging AI Models
In the present legal landscape, characterised by intricate legal frameworks and dynamic regulatory shifts, the effective and precise utilisation of legal citations holds paramount importance. Including these citations substantiates legal arguments and elevates the objectivity and fairness of judicial judgments. Legal citation prediction is a pivotal component of legal research, assisting lawyers and legal professionals identify pertinent legal authorities while formulating legal opinions. In this talk, I aim to scrutinise the most recent AI models employed for legal citation prediction.
In our recent and emerging work, we introduce the GPT2-Small and GPT2-XL models, subjecting them to fine-tuning specific to the citation prediction domain. We also incorporate innovative techniques, such as parameter-efficient fine-tuning (PEFT) and Low-Rank Adaptation of Large Language Models (LORA), tailored to enhance the computational efficiency of the GPT-2-XL model. Furthermore, we introduce a novel approach that harnesses prompts to augment the citation generation process of the GPT-2-XL model.
Joemon Jose is a Professor in the School of Computing Science at the University of Glasgow in Scotland. His past scholarly activities encompass a comprehensive interest in information retrieval, spanning theory, experimental methodologies, evaluation frameworks, and practical applications, all within textual and multimedia domains. At present, his research endeavours are advancing recommender techniques, with a particular emphasis on the intricate exploration of Deep Learning methodologies and Large Language Models (LLMs). He is a programme co-chair for CHIIR’24.