Using Genres to Improve Search Engines

Modern search engines are typically queried with keywords, which foremostly convey the topic of the sought web page. Consequently the resulting top hits are often topically relevant, but nonetheless not what the user wants. The premise of this paper is that the relevance of the hits can be improved when also searching by genre, classification criterion orthogonal to topic. To this end a genre classifier was built using machine learning methods. It was used in web page retrieval to filter out the hits not belonging to the desired genre. This approach considerably improved the relevance of the top ten hits, which indicates that genre classifier can be a useful addition to search engines.