Automatic discovery of new entities using graph reconciliation
Topics: Data Mining, Entity based search, Knowledge Graph, Semantic Search
The patent discusses a method for identifying potential entities from web-based sources by generating a source data graph for each potential entity found in a text document. These graphs are clustered by entity name and type, and through a process of iterative splitting into subclusters, reliable facts about the entities are identified and used to enhance a target data graph. This process is designed to automate and scale the addition of new entities and facts into large data graphs.