University of Hawaiʻi at Mānoa experts have made significant contributions to artificial intelligence (AI) research, with three recent publications highlighting new approaches to information processing that could benefit fields such as healthcare, education and technology.
Authored by experts in the ALOHA Lab, led by Department of Information and Computer Sciences Assistant Professor Haopeng Zhang, these studies advance biomedical text analysis, automatic text summarization generalization, and the evolution of text summarization methods, offering impactful solutions for improving how AI processes and interprets complex information.
“UH Mānoa and the ALOHA Lab are at the forefront of AI research, driving innovations that improve information accessibility, enhance healthcare insights, and adapt technology to meet the unique needs of diverse communities,” said Zhang. “As AI continues to evolve, contributions like these will be instrumental in shaping how technology processes and understands human language.”
The first study, “A Structure-aware Generative Model for Biomedical Event Extraction,” presents a new AI model that improves how computers understand complex medical events in text. Older models struggle to identify relationships between medical terms, especially when events are layered or interconnected. The new model, GenBEE, addresses this challenge by using structured prompts to better recognize and organize important medical information. By applying this model to benchmark datasets, researchers demonstrated its superior performance in extracting critical biomedical information. The paper will be presented at the 30th International Conference on Database Systems for Advanced Applications in Singapore.
Another study, “DomainSum: A Hierarchical Benchmark for Fine-Grained Domain Shift in Abstractive Text Summarization,” explores how AI summarization models handle different types of content. Many current models work well on one specific topic but struggle when summarizing information from new areas. DomainSum evaluates how well these models adapt by categorizing topic shifts into three categories: genre, style and subject. By analyzing existing AI models, this research provides insights on making summarization tools more flexible and accurate across different fields. This research will be presented at the 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics in Albuquerque, New Mexico.
The third publication, “A Systematic Survey of Text Summarization: From Statistical Methods to Large Language Models,” provides an in-depth review of the evolution of text summarization techniques. The study traces developments from early statistical methods to the latest advancements in large language models, highlighting key datasets, evaluation metrics, and emerging trends. As a comprehensive resource for researchers, this survey identifies challenges and proposes future directions for AI-driven summarization. The paper has been accepted by the Association for Computer Machinery Computing Surveys, one of the most influential journals in the field.
