Unsupervised Acquisition of Desktop Application Taxonomies

This paper presents a novel approach for automatically extracting taxonomies of desktop applications from descriptive eventlogs, unobtrusively collected from the user´s computational environment. By using a linguistically inspired clustering technique on application transitions we are able to generate groups of applications that are similar towards their purpose. Attributing the application category (e.g. office application) to a desktop application in an unsupervised manner without manual user engagement yields advantages in many different application areas such as process-oriented workplace-embedded (e-)learning systems or performance management and evaluation.