Author: Can Özgür Yilmaz
Supervisor: Allan Hanbury, Co-Supervisor: Julia Neidhardt
Abstract
In today’s world, usability testing is an essential aspect of the software developmentprocess. Conventional usability testing however is very time and resource consuming,because there needs to be a real person observing the user and trying to discover usabilityproblems from their observations. As a solution to this problem, we present a framework for semi-automated usability testing, in which the user tests the application without an observer, and then their interactions are captured, saved and then analysed to discoverpotential usability problems. Our proposed framework combines four general semi-automated usability test approaches into one, these are: metric based, pattern matching,task based and inferential. To evaluate our proposed framework we developed a semi-automated usability tool (SAUTO) and conducted manual and semi-automated usabilitytests during the development of a web application, the DYLEN tool. We evaluated the framework by comparing the results of manual usability tests to semi-automated usabilitytests and also by conducting qualitative interviews with the developers of DYLEN and an external usability expert. The results indicate that while semi-automated usabilitytesting cannot replace manual usability testing, it can be very useful to carry out both together. Additionally, semi-automated usability testing allows to test a software with a high amount of users which is not possible with manual usability testing. Finally it allows carrying out usability tests after the launch of a software.