Event-based Modeling and Analysis of Web Database Usage Patterns
Oral Defence Date:
Prof. Murphy, Prof. Singh, Prof. Petkovic
Large multimedia websites pose a significant challenge for comprehension of the complex and diverse user behaviors they sustain. This paper presents a novel multi-perspective approach for usability analysis of the large websites. The approach improves upon the state-of-the-art both conceptually and technically. On the conceptual side it brings together information from both the web usage log and web content and structure. On technical side this technique improves upon the state-of-the-art with following features: (1) capturing the semantics of text and image media for static and dynamic web content; (2) recognizing that information goals are defined by emergent user behavior, not statically declared by web design; and (3) session visualizer that provides valuable insight to the usage pattern. Given a user session, the model accurately predicts user information goals and presents them as a list of most relevant words and images. Also, given a set of information goals, the model predicts user navigation as network flow model with a semantically-dependent flow distribution. Differences between predicted optimal and observed user navigation highlights points of sub-optimal website design. The proposed technique is evaluated with real data from the multimedia websites of the Skyserver Project, and the Biocomputing and Media Research Group at San Francisco State University. The experimental evaluation demonstrates the effectiveness of the system in comparison to other similar approaches.
web usability, semantics, media modeling, information foraging, visualization, mining, multimedia analysis, information scent, semantic entropy