Towards automatic semantic annotation of data rich Web pages

Abstract

Semantic annotation is a preliminary step towards the Semantic Web. In the Web scale, the large amounts of heterogonous data sources make manual annotation impracticable; automation is therefore unavoidable. Our work aims to develop a solution for automatic semantic annotation. We focus on data-rich web pages, i.e., pages that contain a list of records. All these pages are related to a particular domain. Most of data-rich web pages are generated automatically according to a template. Many solutions, based on wrapper induction techniques have been established to automatically extract data contained in this kind of pages. The solution we propose follows the data extraction phase. It transforms the data into a set of RDF assertions expressed in accordance with an ontology of the domain of interest. Our approach is based on instances. This means that starting from a set of instances we try to find the concept or property of the ontology that has the best match. We aim here to present the details of our approach and discuss the results produced in the primary experiments which turn to be promising.