A Four Layer Bayesian Network for Product Model Based Information Mining

Business and engineering knowledge in AEC/FM is captured mainly implicitly in project and corporate document repositories. Even with the increasing integration of model-based systems with project information spaces, a large percentage of the information exchange will further on rely on isolated and rather poorly structured text documents. In this paper we propose an approach enabling the use of product model data as a primary source of engineering knowledge to support information externalisation from relevant construction documents, to provide for domain-specific information retrieval, and to help in re-organising and re-contextualising documents in accordance to the user’s discipline-specific tasks and information needs. Suggested is a retrieval and mining framework combining methods for analysing text documents, filtering product models and reasoning on Bayesian networks to explicitly represent the content of text repositories in personalisable semantic content networks. We describe the proposed basic network that can be realised on short-term using minimal product model information as well as various extensions towards a full-fledged added value integration of document-based and model-based information.