Handling Cyclic Axioms in Dynamic, Web-Scale Knowledge Bases

In recent years, there has been an explosion of publicly available Semantic Web data. In order to effectively integrate millions of small, distributed data sources and quickly answer queries, we previously proposed a tree structure query optimization algorithm that uses source selectivity of each query subgoal as the heuristic to plan the query execution and uses the most selective subgoals to provide constraints that make other subgoals selective. However, this constraint propagation is incomplete when the relevant ontologies contain cyclic axioms. Here, we propose an improvement to this algorithm that is complete for cyclic axioms, yet still able to scale to millions of data sources.