Prof. Peter M.D. Gray
Stuart Chalmers
Dr. Kit Hui
Former Group Members:
Dr. Suzanne Embury
Dr. Graham J.L. Kemp
We have been very successful in applying object database technology to the storage of large amounts of real world knowledge, so that it can be combined with intensional and semantic knowledge expressed in Prolog, and also make use of AI search techniques combined with database indexing. This is a natural extension of the frame-based knowledge storage used in expert systems, but it gives the added advantage of secure persistent storage, combined with the typed discipline of using a data model.
The work has been supported by a sequence of EPSRC grants, and is being applied to important problems in molecular biology and geology. In the area of bioinformatics, it has been applied to the storage of protein structure data for use in homology modelling. This led on to combining sequence and structure data for modelling antibodies, and collaboration with industrial and academic groups to suggest structural reasons for observed differences in binding affinity and to guide protein engineering work in developing antibodies with therapeutic and other applications; this was chosen for presentation in an EPSRC brochure.
We are actively pursuing the use of multi-database technology to provide integrated access to heterogeneous, distributed biological databases, including the rapidly expanding protein databases at EMBL (Heidelberg) and EBI (Cambridge). We are helped in this by collaboration with Prof T. Risch (Uppsala, Sweden) through our participation in the EU Active Database Net of Excellence (ACTNET) and through the exchange of personnel.
EPSRC is supporting work on the storage of integrity constraints as declarative semantic knowledge, which can be transformed for a variety of uses, especially in design, based on a knowledge of parts and how to use them. Further, our Prolog representation of constraints lends itself to constraint logic programming techniques (CLP) and we are actively pursuing these as a form of mobile knowledge for Agents.
EPSRC and BT supported the KRAFT project which started in May 96 and which ran for three years. Aberdeen led a collaboration with other database groups at Cardiff and Liverpool and Martlesham to investigate the fusion of constraint knowledge from distributed sources, and its potential for knowledge re-use. Knowledge fusion and re-use are recognised internationally as important problems, and the KRAFT consortium funding consequently approached #1M (of which Aberdeen's share was 222K). This Knowledge Reuse work is being continued with funding from EPSRC for the AKT consortium. BT is also supporting work on the CONOISE project(2002), which takes KRAFT work further in using negotiating agents.
P.M.D. Gray, Large Databases and Knowledge Re-use, in Computing Tomorrow, I.Wand and R.Milner (eds.), Cambridge Univ. Press, 110-126 (1996).
pgray@csd.abdn.ac.uk