A Comparison of GUI and VUI Testing

Software testing is a critical step in the software development lifecycle, and it is equally important to evaluate the graphical user interface of an application. While there are many testing techniques used in testing graphical user interfaces, the number of techniques and studies for voice user interface testing is minimal. This paper presents an overview of selected techniques used in graphical user interface testing, which highlights that automated test case generation offers many benefits to the testing process. Voice user interface testing methods are also reviewed, and while there are some similarities to graphical user interface testing techniques, voice user interfaces do require different types of interaction. A comparison between the two testing groups is made along with the conclusion that the many methods of test case generation used for graphical user interfaces can be applied to voice user interfaces. This paper also proposes how a newly created graphical user interface and a newly created vocal user interface could benefit from selected testing methods presented within the paper. As Harris [9] points out, anyone who needs a justification for usability testing of an application requiring human interaction might want to change careers. Graphical User Interfaces (GUIs) are pervasive in software applications; it is the primary interface a user has with the software. 15 years ago, developers were dedicating over 50% of code to the GUI [18], and they require a great deal of testing. GUI testing, however, is not as pervasive. There are two categories of GUI testing: usability testing and functional testing [2]. Functional testing consists of four types: GUI system testing, regression testing, input validation testing, and GUI testing (does the software work) [2]. GUI testing in this paper refers to primarily the fourth type, but regression testing is touched upon. Often in GUI testing, not all event-sequences are adequately tested [18]. A sample of test cases may be selected, but one sample may produce one result and a second sample may produce a different result, thus, missing software faults [16]. To complicate testing further, new modalities have been introduced to user interfaces, such as voice.