Language-Independent Text Parsing of Arbitrary HTML-Documents. Towards A Foundation For Web Genre Identification

This article describes an approach to parsing and processing arbitrary web pages in order to detect macrostructural objects such as headlines, explicitlyand implicitly-marked lists, and text blocks of different types. The text parser analyses a document by means of several processing stages and inserts the analysis results directly into the DOM tree in the form of XML elements and attributes, so that both the original HTML structure, and the determined macrostructure are available at the same time for secondary processing steps. This text parser is being developed for a novel kind of search engine that aims to classify web pages into web genres so that the search engine user will be able to specify one or more keywords, as well as one or more web genres of the documents to be found.