Dr. Yangqiu Song, PhD Student Wei Fan, and Haoran Li Received Outstanding Paper Award at EMNLP 2024

In EMNLP 2024, PhD student Wei Fan, Haoran Li, and Dr. Yangqiu Song, and their team, received the Outstanding Paper Award for their co-authored paper "GOLDCOIN: Grounding Large Language Models in Privacy Laws via Contextual Integrity Theory".

The award-winning paper proposes GOLDCOIN, an innovative framework designed to ground large language models (LLMs) in privacy laws by leveraging the theory of contextual integrity. GOLDCOIN addresses the critical challenge of aligning LLMs with nuanced legal statutes, such as HIPAA, by generating synthetic legal cases integrating key contextual features in privacy violations. They evaluated the framework on two judicial tasks— determining the applicability of privacy laws and assessing compliance with these laws. The experimental results demonstrate that instruction-tuned LLMs using the GOLDCOIN-generated datasets outperform baseline models significantly, particularly in recognizing privacy risks and improving legal judgment accuracy. Future work involves expanding the framework to other privacy laws, such as GDPR, and enhancing the diversity and realism of generated case studies.

Empirical Methods in Natural Language Processing (EMNLP) is a leading conference focused on natural language processing (NLP) research. Organized by the Association for Computational Linguistics (ACL), EMNLP focuses on research that emphasizes empirical approaches to NLP, including machine learning, statistical modeling, and large-scale data-driven methods. As a premier venue in the field, EMNLP is a research forum for advancements in all these related areas, bridging contributions from academia and industry.

Congratulations again to Dr. Song, Wei, and Haoran for their impactful work!

For more details, please visit the EMNLP 2024 website and access the full paper.