Date Added: Apr 2011
The authors present methods for annotating data with the time when it was learned and for answering queries according to what was known at any point in time. Specifically, they present an RDF knowledge representation that associates facts with their transaction times, and a query mechanism that transforms a time-agnostic SPARQL query and a point in time into a new, time-sensitive query. The transformed query yields the subset of the results of the original query that were valid at the indicated time. In addition, the methods presented here enable non-destructive merging of coreferences. These techniques apply broadly to storage and retrieval systems that require time-based versioning of data and are essential for maintaining temporal perspective in rapidly-evolving analytical environments.