The central theme of our work is the transparent, integrated usage of resources in open, heterogeneous, dynamic environments. The ultimate goal of this research is to build systems that optimally support users in achieving their goals without the need of human intervention.
We approach this subject from two different angles:
First, we investigate systems built upon the paradigm of service-oriented computing. While dealing with the integration of information has a long tradition in the database community, the provision of integrated access to functionality offered in heterogeneous coalitions of autonomous entities has only recently gained interest. The service-oriented computing paradigm provides a basis for these developments. The success of web services shows the potential of this paradigm. However, web services are just a first step on the way to obtaining automatic integration. Important research issues remain to be addressed:
- How can services be described in such a way that it becomes
possible to automate service discovery? Here, as part of the DIANE
project we have developed and are continuing to extend a semantic
service description language, DSD. With this language we take part in
the Semantic Web Services Challenge, where we obtain excellent
- How can services be combined to achieve higher-level functionality? Again, DSD proves to be a good basis for service composition.
- How can approaches to semantic service discovery be evaluated?
While there exists a large number of approaches to semantic service
description, surprising little effort has been spent on evaluations,
yet alone comparative evaluations of the competing results. In the Opossum project, we are
working towards standardized benchmarks in this area.
- How can autonomous entities be motivated to cooperate? Reputation
systems and incentive schemes can be used here and have been
investigated within DIANE.
Currently, we are looking at the combination of service disvovery and
reputation systems and at the application of reputation systems in a
number of application areas including emergency management.
- How can services be executed efficiently? Solutions can be based on query planning and execution mechanisms known from distributed DBMS. Details on our replication/hibernation approach can be found in the DIANE publications.
Second, in cooperation with IBM, we are working on establishing next-generation portals. Portals were originally established to provide a uniform and integrated access point to information and functionality spread over an organization. Nowadays, however, typical portals contain so much information that they are no longer able to efficiently support users. What is needed are therefore portals that adapt automatically to the user and his or her current context. In our Minerva project, we combine semantic web techniques with methods from data mining and approaches developed in the Web 2.0 context to build such portals. Additionally, we aim at supporting ordinary users in developing their own portal applications based on the portlet standard JSR286.
Once we have found satisfactory answers to all of these questions, it will be possible to dynamically combine resources and to dynamically adapt and recommend content in such a way that the human user is supported optimally - maybe without even being aware of it.