PLATAL - A Tool for Web Hierarchies Extraction and Alignment

Abstract

This paper presents PLATAL, a modular and extensible tool for extraction of hierarchical structures from web pages which can be automatically aligned and also manually edited via a graphical interface. Evaluation of alignments can be carried out using standard measures. Web sites are rich sources of information for a range of applications. Tools for automatically extracting structured content from these sources and for comparing content across web sites are valuable resources. For helping in these tasks, we propose PLATAL (Platform of Alignment), a modular and extensible tool that provides an integrated environment for extraction of web hierarchies and alignment creation, edition and evaluation. The main motivation behind PLATAL is to assist users in the complete alignment cycle of two web hierarchies. Differently from other matching tools o ering a visual environment, like OLA [1], Prompt [3], Homer [5], Yam++ [2] and SOA-based tool [4], PLATAL o ers novel functionalities: the possibility of automatically extracting hierarchical structures from the web together with a centralised visual tool for alignment manipulation.

Introduction

PLATAL modules

PLATAL is a standalone tool composed of four modules: ( 1 ) hierarchy extraction module, which extracts fragments from HTML pages using XPath expressions; ( 2 ) automatic alignment module, which implements a set of terminological (pre x, su x, edit-distance) and structural matching techniques (similarity of parents and children entities) for generating equivalence correspondences; ( 3 ) manual alignment module, which allows users to edit or create alignments; and ( 4 ) evaluation module, which takes two alignments and computes precision, recall and F-measure measures. These modules operate independently of each other and alternative implementations can be added instead. Figure 1 shows a screenshot of automatic alignment creation. After loading two hierarchies, each hierarchy will be displayed in the respective section. Then, users can select one or more alignment processes and start them (`Start Alignment Process'). If at least one method founds one correspondence between two entities, the user can see it by

Bernardo Severo, Cassia Trojahn, and Renata Vieira selecting the source or target entity in the hierarchies ( eld `Correspondences'). Alignments can be exported in the Alignment format3 (`Save').

Conclusions and future work

We have presented a visual tool for extraction, alignment and evaluation of web hierarchies. To the best of our knowledge, there is no publicly available environment integrating all these features together. As future work, we plan to improve the visualisation of alignments, develop a web-based version, allow parametrisation and customisation of alignment techniques through the user interface, and add a multilingual ontology matching module. 3 http://alignapi.gforge.inria.fr/format.html

References

1. J. Euzenat , D. Loup , M. Touzani , and P. Valtchev . Ontology Alignment with OLA . In 3rd EON Workshop , pages 59 { 68 , 2004 .

2. D. H. Ngo and Z. Bellahsene . YAM+ + : (not) Yet Another Matcher for Ontology Matching Task . In BDA, France, 2012 .

3. N. F. Noy and M. A. Musen . PROMPT: Algorithm and Tool for Automated Ontology Merging and Alignment . In 17h AAAI , pages 450 { 455 , 2000 .

4. K. W. Onn , V. Sabol , M. Granitzer , W. Kienreich , and D. Lukose . A visual soabased ontology alignment tool . In OM, 2011 .

5. O. Udrea , R. Miller , and L. Getoor . Homer: Ontology visualization and analysis . In Demo session ISWC , 2007 .