MAC/FAC Retrieval of Semantic Workflows
Ralph Bergmann, Alexander Stromer · The Florida AI Research Society · 2013
This paper presents a novel two-step retrieval method for semantic workflow cases, inspired by the MAC/FAC (“many are called, but few are chosen”) model proposed by Gentner and Forbus. MAC/FAC retrieval is motivated by the computational complexity of graph matching, which is usually involved in the similarity-based retrieval of workflows. An additional computationally efficient retrieval step (MAC stage) is introduced prior to the graph-based retrieval (FAC stage) to perform a pre-selection of potentially relevant cases. The MAC stage is based on a feature representation of the workflows automatically derived from the original graph-based representation. In the paper, we briefly introduce previous work on the semantic workflow retrieval and then we describe the pre-selection step in more detail. A comprehensive evaluation with case bases from the cooking domain is reported with demonstrates that the retrieval time can be significantly reduced without significant negative impact on the retrieval quality.