Uncovering functional dependencies in MDD-compiled product catalogues

Product catalogues are usually represented as tables. They are regularly updated with new variants and, over time, various forms of inconsistencies or undesirable properties can be introduced, especially when global changes are made. We argue that by compiling product catalogues into decision diagrams we can support a number of high-level queries for detecting and checking for various forms of inconsistency, as well as verifying other properties relevant for user interaction. In particular, a class of functional dependency-based inconsistencies can be detected efficiently. An example is verifying properties such as: "each identical configuration of a product has the same price". This paper presents a number of algorithmic advances. We illustrate the usefulness of our approach by evaluating these high-level properties over realworld publicly available product catalogues.