The World Wide Web can be considered an infinite source of information for both individuals and organizations. Yet, if the main standard of publication on the Web (HTML) is quite suited to human reading, its poor semantics makes it difficult for computers to process and use embedded data in a smart and automated way.
In this paper, we propose to build a bridge between HTML documents and external applications by means of so-called mapping rules. Such rules mainly record a semantic interpretation of recurring types of information in a cluster of similar Web documents and their location in those documents. Relying on these rules, HTML-embedded data can be extracted towards a more computable format. The definition of mapping rules is based on direct user input mainly for the interpretation part, and on automatic computing for the location of data in HTML tree structures. This approach is supported by a user-friendly tool called Retrozilla.