Learning relational facts from the web: A tolerance rough set approach

•A key issue when mining web information is the labeling problem.•Tolerance rough sets are used to structure categorical instances and relations.•We report a semi-supervised algorithm (TPL) that labels the ontological information.•The Never Ending Language Learner (Nell) system provides the ontology.•The performance of TPL compares well with CBS and CPL and handles concept drift.