Automatic Generation of Ontologies from Canonicalized Web Tables

Ontology creation is a daunting task—manual creation is tedious and time consuming, and automatic creation is disappointingly inaccurate. But for applications such as the semantic web or making web content directly queriable, we must facilitate ontology creation, making it reasonable to produce the vast number and variety of ontologies re- quired for future web applications. In this paper we describe a tool devel- oped to automate the generation of ontologies from ordinary web tables. The process is akin to reverse-engineering relational tables to concep- tual models, but must account for a much greater variety of table layout patterns. The tool uses auxiliary knowledge and heuristic rules to select, enhance, and modify ontology elements discovered in tables. Although designed to operate fully automatically, the tool allows users to intervene to direct and correct when necessary so that they can always acquire a satisfactory ontology from a given web table. Experimental evaluations show that the automatic ontology acquisition process can perform well, yielding F-measures of 90% for concept recognition, 77% for relationship discovery, and 90% for constraint discovery in web tables selected from the geopolitical domain.