Automatic extraction of structurally coherent minitaxonomies

Abstract. In this paper we demonstrate an automatic approach for emergent semantics modeling of ontologies. We follow the collaborative ontology construction method without the direct interaction of domain users, engineers or developers. A very important characteristic of an on-tology is its hierarchical structure of concepts. Semantic web is heavily dependent on the XML paradigm, which inherently follows the hierarchi-cal structure. We consider large sets of domain specific schemas as trees and apply frequent sub-tree mining for extracting common hierarchical patterns. Our experiments show that these hierarchical patterns are good enough to represent and describe the concepts of the domain ontology. The technique further demonstrates the construction of the taxonomy of domain ontology. In this regard we consider the largest frequent tree or a tree created by merging the set of largest frequent sub-trees as the taxonomy. We argue in favour of the trustabilty for such a taxonomy and related concepts, since it has been extracted from the schemas being used with in the specified domain