Answering Threshold Queries in Probabilistic Datalog+/– Ontologies
The recently introduced Datalog+/– family of ontology languages is especially useful for representing and reasoning over lightweight ontologies, and is set to play a central role in the context of query answering and information extraction for the Semantic Web. Recently, it has become apparent that it is necessary to develop a principled way to handle uncertainty in this domain. In addition to uncertainty as an inherent aspect of the Web, one must also deal with forms of uncertainty due to inconsistency and incompleteness, uncertainty resulting from automatically processing Web data, as well as uncertainty stemming from the integration of multiple heterogeneous data sources. In this paper, we take an important step in this direction by developing the first probabilistic extension of Datalog+/–. This extension uses Markov logic networks as underlying probabilistic semantics. Here, we especially focus on scalable algorithms for answering threshold queries, which correspond to the question “what is the set of all atoms that are inferred from a given probabilistic ontology with a probability of at least p?”. These queries are especially relevant to Web information extraction, since uncertain rules lead to uncertain facts, and only information with a certain minimum confidence is desired. We present two algorithms: a basic approach and one based on heuristics that is guaranteed to return sound results.
Unable to display preview. Download preview PDF.