Semantic data models and business context modelling

In maintaining a large data model, it is frequently necessary to determine why a particular data item is in the model, and whether it is still relevant to the business or activities that the model is intended to support. When the model is first built, the authors usually have an understanding of how the data items relate to the underlying business requirements. However, that mapping of data items to business requirements or business context has frequently not been a part of the information that is captured, and the knowledge of the business relevance of particular data items is lost as the original authors move on to new projects or new companies.

As the proportion of erroneous, ambiguous, or irrelevant data items increases in a data model, the value if the data model is diminished. Without a knowlege of the business applicability of each data item, model maintainers cannot easily perform the necessary maintenance on the model to keep it relevant and usable.

Business context models differ from many data models in that they contain “instance” information and not just “data type” information. For example, a data model may contain a type/field for representing a country, but a business context model will need to contain actual identifiers for particular countries. The ability to capture detailed instance information and integrate it with data type information is what tends to separate so-called “knowledge” or “semantic” technologies from simpler data modelling technologies.

Some companies and organisations are now looking to use semantic technologies like RDF and OWL for business and context modelling, in spite of the deployment and training costs of introducing these into their existing technology mixes. This presentation looks at the work of a pair of ISO and UN/CEFACT working groups that are working to integrate semantic business context modelling into their existing message-modelling and XML-schema-generation methodologies. The intent of these groups, at least at this initial stage, is not to replace the existing modelling technologies, but to augment them. The question of how the information can be kept synchronized across the technology boundary between the semantic and non-semantic models will also be discussed. This is an approach that can be applied to a wide range of existing data models without the expense of rebuilding the existing models from scratch using a new technology base.