Reverse engineering ontologies from performance systems

Considerable effort is associated with the development, validation and integration of ontologies. This paper suggests that an alternative, or possibly complementary approach, to engineering ontologies is to retrospectively and automatically discover them from existing data and knowledge sources in the organization and then to combine them if desired. The method offered assists in the identification of similar and different terms and includes strategies for developing a shared ontology. The approach uses a data analysis technique known as formal concept analysis to generate an ontology. The approach is particularly strong when used in conjunction with a rapid and incremental knowledge acquisition and representation technique, known as ripple-down rules. However, any data that can be converted into a crosstable (a binary decision table) can also use the approach. The ontological representation is not as rich as many others but we have found it useful for uncovering higher-level concepts and structure that were not explicit in the performance data. If richer models are required our approach may provide a quick way of developing a first draft and gaining initial ontological commitment.