The main objective of web genre categorization is project is to improve genre based navigation of a collection web documents by introducing genre as a parameter that can be attached to a web document and be automatically classified.
One of the challenges of in integrating genre as a navigational tool is finding the right model of taxonomy that satisfies the complex communication pattern presented. Through trends in web such as web directories and collaborative tagging we have found that overlapping between categories in both genre and domain as one emerging characteristic of digital communication.
In machine learning data with overlapping categories such as the above are generally associated with multi-label problem, a special case of learning where instances in the data are allowed to be mapped to more than one class.
In this presentation we will address the aforementioned characteristic and discuss various aspects of multi-label learning as one possible model for learning genre.