A framework for web table mining

Abstract

Web table mining is about information extraction from tables published inside web pages as HTML texts. Most previous work on this subject makes use of the tags to discover components of the table. Our work treats web as a distinct publication media, in two ways. We argue that new types of table format have been developed specially for the web. We also argue that the visual cues embedded within the HTML text, are utilized by the authors to direct the viewer on how to read the contents contained a web table properly. We develop a framework for comprehensively analyzing the structural aspects of a web table, within which rules are devised to process and extract attribute-value pairs from the table. This approach to web table mining is validated by good experimental results.