The Software Therapist: Usability Problem Diagnosis Throught Latent Semantic Analysis

We propose an unprecedented suite of Usability Engineering software tools to be built upon the conceptual foundation of Virginia Tech's User Action Framework (UAF). We will use K-A-T's proprietary Latent Semantic Analysis (LSA) methods and software tools to validate and refine the UAF. We will also use LSA as the underlying analysis engine for the software tools, which will provide support for problem reporting, usability problem extraction, analysis, diagnosis, plus links to related literature and prescriptive solutions. Building upon Phase I tool prototypes, commercial grade development of the suite is proposed for Phase II. The major research thrust is exploration of LSA techniques for free text analysis, trained on the literature of Usability Engineering and a significant library of usability problem reports, to validate and tune the taxonomic structure and content of the UAF. Because no standardized vocabulary exists for usability engineering, simple keyword methods cannot reliably classify problem reports. In contrast, LSA can provide highly reliable measures of semantic similarity between texts even when there is no keyword overlap. In Phase II we will evaluate the developed software tools in usability labs and courses at Virginia Tech, the US Air Force Academy, and commercial software development operations.