Factors that influence query reformulations and search performance in health information retrieval: A multilevel modeling approach
本研究采用多层建模方法，分析了系统特征和用户教育背景对查询重构类型的影响，以及查询重构类型与搜索绩效的关系，为理解健康信息检索中的用户行为提供了新视角。
Query reformulations can occur multiple times in a session, and queries observed in the same session tend to be related to each other. Due to the interdependent nature of queries in a session, it has been challenging to analyze query reformulation data while controlling for possible dependencies among queries. This study proposes a multilevel modeling approach in an attempt to analyze the effects of contextual factors and system features on types of query reformulation, as well as the relationship between types of query reformulation and search performance within a single research model. The results revealed that system features and users' educational background significantly influence users' query reformulation behaviors. Also, types of query reformulation had a significant impact on search performance. The main contribution of this study lies in that it adopted the multilevel modeling method to analyze query reformulation behavior while considering the nested structure of search session data. Multilevel analysis enables us to design an extensible research model to include both session‐level and action‐level factors, which provides a more extended understanding of the relationships among factors that influence query reformulation behavior and search performance. The multilevel modeling used in this study has practical implications for future query reformulation studies.