Mining the Web through verbs: A case study

Abstract

Mining non-taxonomic relations is an important part of the Semantic Web puzzle. Building on the work of the semantic annotation community, we address the problem of extracting relation instances among annotated entities. In particular, we analyze the problem of verb-based relation instantiation in some detail and present a heuristic domain independent approach, based on verb chunking and entity clustering, which doesn't require parsing. We also address the problem of mapping linguistic tuples to relations from the ontology. A case study conducted within the biography domain demonstrates the validity of our results in contrast to related work, whilst examining the complexity of the extraction task and the feasibility of verb-based extraction in general. © Springer-Verlag Berlin Heidelberg 2007.