I was involved in various other ESPRIT and DFG projects.
My own view of the semantic Web is the following: currently programmers program in the context of an operating system. They can use the resources of the operating system (kernel functions, libraries, other programes etc.). In a semantic web it should be possible to program in the context of the whole world-wide-web, and use the resources of the Web. An analog to kernel functions of an operating system would be Web services which provide basic functionalities needed by many applications. Currently there are only very few such services used widely. One is the Domain Name Service, mapping domain names to IP addresses. Advanced Web applications could benefit enormously if the Web itself would provide more such functionalities.
I started the development of a "temporal notions" service. It consists of several levels. The lowest level provides datastructures and algorithms for time points and crisp and fuzzy temporal intervals. The next level provides a representation of periodic temporal notions as labelled partitionings of the real numbers. Calendar systems are modelled as a particular combination of certain labelled partitionings. The third level is the GeTS language (GeoTemporal Specifications), which can be used to specify and use application dependent temporal notions (`my weekend', `my working hours', `Jane's birthday', `the next three days', `before Christmas' etc.) Finally, the fourth level would integrate these temporal notions in other systems, for example a Description Logic.
More details can be found at the CTTN, homepage (Computational Treatment of Temporal Notions).
The main advantage of logic as a knowledge representation language is its declarativity. This means in particular that a logical specification is usually easy to obtain, easy to read and easy to change. The disadvantage is that the specifiation and solution of many problems require specific data structures and algorithms which are very difficult to formulate and reason with in logic. I am therefore working on combinations of logical and non-logical languages and algorithms. The goal is to develop languages and inference methods which use as many special purpose datastructures and algorithms as possible, but still have the nice declarative features of logics.
The following combination approaches have been investigated:
Last Modified: Sunday, 08-May-2005 16:40:41 CEST