Extração e alinhamento de hierarquias em páginas WEB

Abstract

The heterogeneity of the ways information is presented on the web is a characteristic which complicates the analysis between different sources. Even in hierarchical structures, which have a minimum relation of order, there is no standard for how to display the elements and how to reference them. Therefore, this work s main focus is to present a visual and extensible tool that centralizes and supports operations on such structures in web pages. To that end, the PLATAL (Platform of Hierarchy Extraction and Alignment) tool was developed, to facilitate the various operations of hierarchy alignment. The tool has four main modules: one for extracting hierarchies of web pages, making them available for manipulation in standard formats of the semantic web; one for automated alignment of these hierarchies, based on various heuristics and ontology alignment techniques; one for manual alignment of hierarchies, allowing the creation of reference alignments; and finally, one for evaluation of alignments, through the analysis of precision and recall. To evaluate the heuristics of alignment, experiments were performed in the field of e-commerce. The results were compared with that produced by other tools described in the literature. Therefore, this work contributes as a way to enable the creation of aligned hierarchies from heterogeneous structures found on the web.