Overview of Automatic Genre Identification

Genre is a category of artistic, musical, or literary composition characterized by a particular style, form, or content (according to Merriam-Webster Online Dictionary [2] – among two other, less relevant definitions). It can be debated whether a genre can be defined by style (and form) alone, or must content (topic) also be taken into account. For example, can science fiction be distinguished from all other literary genres only by style or must one also consider that it talks about spaceships and time travel? But for the purpose of information retrieval, topic component of genre can and should be disregarded, because topic is usually treated separately and is characterized by keywords (and other methods), so genre should be orthogonal to topic. This report is written with a general corpus such as WWW in mind. Genres that cannot easily be distinguished by style are probably too fine-grained anyway for a corpus like that: it would be nice to be able to search for science fiction, but at this granularity, there are simply too many genres. It is probably best to limit oneself to genres such as homepage, FAQ, scientific paper, etc. Another common way to describe the genre of a document is through the purpose of the document, but looking for a way to automatically detect the purpose of a document leads back to style.