Presented by Elisa Kendall & Dean T Allemang
Knowledge graph technology enables question answering across data silos at scale. In order to provide an understanding of the data and explain the results returned by complex queries, machine learning, and other applications, a common vocabulary is essential.
Industry-standard models are often criticized due to their complexity, including significant breadth and depth. This is a consequence of the unique position that an industry-level ontology plays with respect to other data models. In short, an industry ontology has to anticipate a wide range of design options in enterprise data models – and include elements that reflect those options – to mediate viewpoints.
How can an industrial ontology manage this complexity and make it usable at an enterprise level? Through modeling patterns that allow for the range of variation required while allowing the model to be understandable and applicable to enterprise data modelers.
We illustrate some of these patterns and this effect with specific examples from FIBO (Financial Industry Business Ontology) mappings to enterprise data models. Lessons learned in developing and using these patterns not only in finance, but for automotive, retail, pharmaceutical, and industrial applications, provide a level of confidence in their general applicability.