Information extraction from HTML pages and its integration

Abstract

We propose a method for transforming HTML tables that have different structures into a common XML list structure and integrating them. This integration enables us to browse and compare all information in separate HTML pages uniformly. This paper focuses on the tasks of information extraction from tables and its data categorization. For this purpose, we compare three algorithms: (I) data classification using a support vector machine, (II) table structure estimation and data categorization using a hidden Markov model, and (III) data classification by the combination of a support vector machine and hidden Markov model. Finally, we report the experimental results.