Minimizing conservativity violations in ontology alignments: algorithms and evaluation