Knowledge Patterns: Working Note 13

In our earlier work, components were considered self-contained logical theories, and composition was performed by simply including these theories in a larger KB, with suitable inference to account for interactions among components. In this working note, we sketch out a more general vision of components, which is more in tune with our original goal that they capture “cliches” or “patterns” of relationships. In this approach, a component’s axioms are transformed, or morphed, before being imported to a KB, by renaming its symbols. We describe how this can be done by manually specifying such renamings, and then how such renamings can be automatically computed by the inference engine, including for potentially unanticipated situations. Finally, we describe how questions can be answered by runtime construction of a “relevant” KB using this approach, rather than by querying a massive, pre-built KB.