Contribution to or ideas on semantic web systems with respect to storage and retrieval
We are developing a system that uses RDF to provide a single access point for corporate knowledge bases for engineering companies. It is a system that supports engineering design methodologies including MOKA, which can be summarised as a methodology for organising information in corporate knowledge bases and setting out a foundation for developing formal models that closely integrate with design activities. We want to provide a way that individual engineering designers can relate their CAD work seamlessly back to their companies accumulated design knowledge.
We can use RDF for organising existing and ad hoc knowledge resources. We can provide custom built, high performance software for accessing document fragments, but RDF will be the key framework for linking this information together. The process we are supporting requires that objects in the business domain are initially informally described, and that certain types of formal detail can subsequently be added into the framework.
However, we need to model information that we cannot fit into RDF. The formal models that we must support include product models with mathematical relationships between components. We also need to be able to express rules that apply to the product model during design processes. This is a case of quickly requiring a more expressive language than any current RDF-based one.
However, RDF is very fitting for part of our purpose, and so we devised way of embedding our additional information in RDF and we have developed our own internal representations of the RDF and engineering constraints so that they can be fully integrated. We need to do this so that we can support our customers in representing engineering constraints that integrate with the knowledge model. We have developed a way of validating both parts of the knowledge model by type checking.
The work we are doing is based on our experience at a major European Aerospace engineering company, and related marketing opportunities that are being developed in other engineering sectors. We are a small team of software professionals with many years of experience in developing IT solutions.
Storage issues
The approach we have taken to storing RDF is largely determined by the methodology that we are trying to support, which essentially starts by identifying and describing business objects individually. We are clustering statements about identified business objects and storing them separately as RDF fragments. At Airbus, we are storing XML ontology fragments relating to individual objects in a relational database, and indexing them by the object names.
However, to validate one ontology fragment, we need to be able to load other related ones, and to do this we have developed a way of incrementally loading ontology fragments into a server process. The problem that we are trying to deal with here is scalability, and we are trying to deploy a solution that is based on lazily loading ontology information. This is something that we want to use as a basis for ontology visualisation software.
At Airbus, we are also deploying a multi-version ontology system. Because of the course granularity or our physical storage mechanism, we can access specific versions of ontology fragments. We do this so that knowledge engineers can work on different versions of ontologies concurrently, in isolation.
This design is not suited to lightweight querying of RDF structures directly from the physical storage structures. Rather, it is suited to incrementally accessing detailed information from RDF fragments. This is because we are primarily trying to manage rich information structures rather than providing a query facility.