Analyzing User Behavior in Scenario-Based Virtual Reality Environments Using Knowledge Graphs

• Published in: 2024 16th IIAI International Congress on Advanced Applied Informatics (IIAI-AAI) pp. 638 - 643
• Main Authors: Taninaka, Kensuke, Furuya, Yoshiki, Sakamoto, Keigo, Mine, Tsunenori, Fukushima, Shogo
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06-07-2024
• Subjects: Informatics; Knowledge graphs; Object recognition; Ontologies; Psychology; Resists; Scenario-based virtual space; Social factors; User behavior; Virtual reality; Vocabulary
• DOI: 10.1109/IIAI-AAI63651.2024.00120

Widespread adoption of head-mounted displays (HMDs) has led to an increase in virtual reality (VR) content experiences. Because a VR space has fewer physical constraints than the real world, users behaviors may differ from that in the real world. Particularly in scenario-based VR spaces where users are required to perform specific tasks, the focus has been on acquiring user behaviors that align with the scenario. However, analyzing the characteristics and degrees of freedom of user behavior that deviate from a given scenario is important for developing and analyzing games and learning content. In this study, we propose a system that converts HMD sensor data into a Knowledge Graph (KG) using a custom ontology to analyze user behaviors and their characteristics in a VR space with tasks but high degrees of freedom. To evaluate the proposed system, we investigated the behavioral characteristics of two users who operated a VR application to learn English vocabulary as an example. Specifically, we generated KG from the operation log data of the objects manipulated by the users in the VR application, which was then used to analyze the differences between the two user characteristics. The results showed quantitative and qualitative differences between users in terms of the degree of compliance in task performance, proportion of interactions with specific objects, and other characteristics. User 1 had 1.8 times as many interactions as User 2. In addition, compared to User1, User2 showed approximately 3 times higher values in terms of the proportion of out-of-scenario behaviors among all behaviors and the time spent on out-of-scenario behaviors.