Title of the presentation: Deriving web genres from text types: a corpus-based approach

Long Abstract: 
From a textual point of view, the web is a huge reservoir of documents. On the web virtually everything can be seen as a ‘document’ or better a ‘web page’. The sheer amount of texts available is just overwhelming. Furthermore, the web is mainly wild and uncontrolled. This becomes clear if we compare a ‘tamed’ resource of the paper world, like the British National Library, and the ‘untamed’ electronic English Web. All the knowledge conveyed by the classification systems of library catalogues is simply nonexistent on the web. 
In this talk, I would like to focus on a specific issue related to classification on the web: the automatic identification of genres of web pages. In particular, I would like to present a model where text types are inferred using a modified version of Bayes' theorem, and web genres are derived combining the inferred text types and other traits (mainly layout and functionality tags) using handcrafted if-then rules.
While in Biber's multidimensional approach text types are inductively-derived textual dimensions providing insight into cross-genre variation, I suggest a method that combines deductive and inductive approaches. The method is deductive because the co-occurrence and the combination of features in text types is decided a priori by the linguist. It is also inductive because the inference process is corpus-based. This method is more flexible and informative than standard classification models. It can also provide useful information about a new genre for which no label is yet available. In a few words, this model combines the classificatory and the descriptive frameworks summarized in Biber (1994: 37).

Short Abstract:
In this talk, I would like to focus on a specific issue related to classification on the web: the automatic identification of genres of web pages. In particular, I would like to present a model where text types are inferred using a modified version of Bayes' theorem, and web genres are derived combining the inferred text types and other traits (mainly layout and functionality tags) using handcrafted if-then rules.