Developing a pandemic-specific taxonomy and artificial intelligence models for classifying information needs during COVID-19

This study develops a pandemic-specific taxonomy and artificial intelligence–based classification models to categorize COVID-19-related information needs expressed by the public. A total of 91,772 questions were collected from a leading social Q&A platform in South Korea and analyzed using Python. From this dataset, 10,000 questions were randomly sampled and manually annotated to construct a training corpus. The annotated data were then used to fine-tune two classification models, KoELECTRA and GPT-4o-mini. The analysis identified 12 primary categories of pandemic-related information needs, encompassing both informational and socio-emotional dimensions. The results demonstrate that artificial intelligence–based models can classify large-scale user-generated health queries with a high level of accuracy, offering a scalable alternative to traditional content analysis approaches. These findings suggest that artificial intelligence–based classification systems may support the organization and analysis of online health discourse by facilitating the identification of emerging information needs and patterns. Such systems may complement existing Library and Information Science tools, by enabling more adaptive and context-sensitive approaches to categorizing health-related information in dynamic crisis environments.