Assessing similarity for case-based web user interface design

Abstract

It has been said “all design is redesign”, and it is particularly true for websites, whose number in the today’s online environment has reached 1 billion. In our paper, we justify case-based approach (CBR) to designing web user interfaces (WUIs) and outline some currently unsolved problems with its application. In this research work, we focus on definition and measurement of similarity, which is essential for all the stages of the CBR process: Retrieve, Reuse, Revise, and Retain. We specify the structure of a case in the web design domain (corresponding to a web project) and outline the ways to measure similarity based on the feature values. Further, we construct artificial neural network model to predict target users’ subjective similarity assessments of websites that relies on website metrics collected by our dedicated “human-computer vision” software. To train the model, we also ran experimental survey with 127 participants evaluating 21 university websites. The analysis of the factors’ importance suggests that frequency-based entropy measure and the proposed index of difficulty for visual perception affected subjective similarity the most. We believe the described approach can facilitate design reuse on the web, contributing to efficient development of more usable websites crucial for the e-society advancement.