Website Structure Visualisation for Supporting User Navigation

Abstract

This thesis investigates which visualisation techniques support fact-finding tasks through the visualisation of website structures. Large websites, such as university websites, often contain hundreds of thousands of pages organised in hierarchies, which hinder users’ ability to locate specific information. Search engines and navigation menus often fail in such contexts, leading to disorientation and inefficient browsing. The thesis begins with a literature review examining web navigation behaviour, structural complexity, and the role of information scent. Fact-finding tasks were identified as the most appropriate task type for evaluating visualisation techniques. The study then conducted web data collection and processing, including crawling three university websites and constructing region-level hierarchical datasets that formed the basis of all evaluations. Two evaluations were conducted. The first was a technical evaluation of twelve visualisation techniques using three metrics: aspect ratio, visualisation area, and label overlap. Squarified treemap, bubble tree, and tree leaf demonstrated the strongest technical performance and were selected for further investigation. The second evaluation consisted of three user experiments. The first experiment examined performance in identifying known-target regions. Results showed that the squarified treemap and tree leaf outperformed the bubble tree, with performance influenced by region order and number of regions. The second experiment examined explanatory navigation. Results showed that the squarified treemap supported better performance, with effects influenced by information scent and number of clicks. The final experiment focused on the squarified treemap and examined how different encodings and structural metrics affected navigation performance. Results showed that saturation encoding, particularly for outgoing links, led to faster task completion times and fewer errors. The findings provide empirical evidence on how the squarified treemap supports efficient fact-finding tasks in website structures.