Supporting ontology management through self-describing concepts

摘要

Lately, several research projects developed more coherent hierarchies, in the form of taxonomies (e-construct project and e-COGNOS project, for example). However, there is a need to bridge the gap between these taxonomies, wrap legacy classification systems, and link the construction industry to other industries. i.e. we need to develop formal clustering mechanisms to allow these (and future) taxonomies to interoperate. This paper presents an architecture for clustering construction concepts. The proposed architecture clusters construction concepts into two main dimensions. The first: Practical Concepts, encapsulates the main entities in construction terms (such as Process, Actor, Resource and Products). The second: Clustering concepts, is meant to capture the attributes, types and constraints of entities. Using genetic algorithms, each concept will carry within itself full reference to its attributes, constraints and other related concepts. Such self-describing concept representation is in valuable in data mining exercise, where formal concept analysis (FCA) and genetic analysis techniques could be used to automatically generate construction taxonomies and/or concept lattice. The clustering concepts are based on the basic IDEF0 dimensions (input, output, mechanisms and controls). As such they contain metadata about concepts. This will allow other industries taxonomies to access concept in any taxonomy based on the proposed architecture.