The Gnowsis Semantic Desktop approach to Personal Information Management
PhD-Thesis, Fachbereich Informatik der Universität Kaiserslautern, ISBN 978-3-86624-449-8, Dissertation.de, Berlin, 6/2009.
We should no longer ask whether we have enough information, we should rather ask if we can manage the information we have. In theory, a person could store and access all information experienced in a lifetime. But to categorize and understand this information, integrated software tools are needed. Since 2003, I have been working on the vision of a Semantic Desktop to create a tool for keeping information. Based on semantic web standards developed by the World Wide Web Consortium (W3C), the information needed by a person to do knowledge work can be integrated and organized. In this thesis, a software architecture for the Semantic Desktop is presented consisting of ontologies, services, and applications. The core ontology is the Personal Information Model (PIMO), a personalized categorization and organisation framework. With it, a formal representation of the mental model of a user can be expressed. This personal information model is then used across applications and across domains, integrating information sources into a coherent view of the world. Existing documents are classified with multi-perspective classification, removing the limitations of hierarchical file structures. Several services are defind for the Semantic Desktop architecture. They are designed to run as operating system extensions and provide functionality to store data, annotate it, and support the user in PIM activities. To provide a user interface, different interface metaphors are tested in various prototypes. These range from small plugins that extend existing software with Semantic Desktop functionality to complex applications that allow generic resource browsing and annotations. The approach was evaluated in end-user experiments in order to find out how PIMO reflects the personal mental model of the user. It was verified to support the users in structuring their documents across applications according to their mental model and in retrieving information based on these structures. The realized implementation shows that the architecture is valid and works in real-life settings. The evaluated prototypes have shown their benefits in many person-years of productive usage. As part of our approach to software engineering at DFKI, we released our source code as free software. Other researchers have used our prototype as a basis for their work and have provided us with valuable feedback. Within the EU research project NEPOMUK, the presented results were integrated into the popular Linux desktop KDE (version 4) and are shipped to millions of users. The Aperture framework, initiated as part of this thesis, was downloaded more than 10.000 times and is in productive use.