Avoiding Alignment-based Conservativity Violations through Dialogue

A number of ontology matching techniques have been proposed that rely on full disclosure of their ontological models prior to the construction of the alignment. However, within open and opportunistic environments, such approaches may not always be pragmatic or even acceptable (due to privacy concerns). Several studies have focussed on collaborative, decentralised approaches to ontology alignment, where agents negotiate the acceptability of correspondences (i.e. mappings between corresponding entities in different ontologies) acquired from past encounters, or try to ascertain novel correspondences on the fly. However, such approaches can lead to logical flaws that may undermine their utility. In this paper, we extend a dialogical approach to correspondence negotiation, whereby agents not only exchange details of possible correspondences, but also identify potential violations to the so-called conservativity principle, where novel but undesirable entailments between named concepts in one of the input ontologies emerge. We present a formal model of the dialogue, and show how \conservativity violations can be repaired (using an existing correspondence repair system) during the dialogue through the exchange of repairs. We then illustrate how agents negotiate over possible correspondences and repairs by means of a walkthrough example.