by Daniel Schall
My PhD centered around the concept of Human-Provided Services. I developed a novel framework allowing people to provide user-defined services based on their capabilities, expertise, and reputation. Additionally, I focused on the definition and implementation of link-based importance analysis algorithms (DSARank - Dynamic Skill- and Activity-aware PageRank).
My PhD advisors were Schahram Dustdar and Frank Leymann
Research Statement (PDF), Thesis (PDF, BIB), Presentation (PDF, SlideShare)
Web-based collaboration platforms evolve into service-oriented architectures by promoting
composite and user-enriched services. In such platforms, the collaborations typically
include both human and software services, thus creating highly dynamic and complex interactions.
However, in existing platforms users cannot specify different interaction interfaces
as services that can be reused in various collaborations. We argue that people need more
ways to indicate their availability and desire to participate in collaborations.
Furthermore, open service-oriented environments require a flexible yet reusable collaboration
model because compositions comprise interactions between people and a number of
software services. The presented work introduces Human-Provided Services (HPS), which
can be included in ad-hoc and process-centric collaborations. The HPS framework fosters
the user-driven integration of human capabilities into service-oriented infrastructures,
thereby promoting reusability and flexibility of interaction flows. By using the framework,
people can manage their interactions and provide services in dynamic collaborations.
Moreover, in open and dynamic collaboration environment, typically a very large number of people collaborate and interact by using different collaboration tools and platforms. It is important to determine expertise and skill level of users. Somebody seeking help or advice on how to solve a specific problem needs to be able to find the right expert. However, the expertise and importance of users changes depending on performed tasks, interactions with other users, as users gain know-how by collaborating with other experts, and based on the information users receive from other people. An expert recommender algorithm must consider the expertís interest in a certain area. For example, a scientist may have done research in a certain field; however, the scientist might change his/her principle research domain over time and therefore no longer be the right expert to contact. Thus, the interest and activity level of a person in a specific field must be considered. We believe that ranking models should not only rely on profiles or skill information that need to be maintained and manually updated by users. It is unlikely that a single skill- or expertise-ontology is sufficient to capture the concepts and requirements of various collaboration domains. Tagging mechanisms can be used to classify information and to derive the context of interactions. Tags provide a) input to derive skills and user-interests; and b) the context of activities and interactions. The challenge is to devise a ranking model that is able to capture these dynamic, context-dependent properties.
Keywords: Human-Provided Services, Crowdsourcing, Web Services, Relevance Ranking, PageRank, Expertise Mining, Social Dynamics
"Human Interactions in Mixed Systems - Architecture, Protocols, and Algorithms", 02/2009.
Back to Daniel Schall's Web page.