Fuzzy Rough Set-Based Unstructured Text Categorization.

In this paper, we have proposed a fuzzy rough set-based semi-supervised learning algorithm (FRL) to label categorical noun phrase instances from a given corpus (unstructured web pages). Our model uses noun phrases which are described in terms of sets of co-occurring contextual patterns. The performance of the FRL algorithm is compared with the Tolerance Rough Set-based (TPL) algorithm and Coupled Bayesian Sets-based(CBS) algorithm. Based on average precision value over 11 categories, FRL performs better than CBS but not as good as TPL. To the best of our knowledge, fuzzy rough sets has not been applied to the problem of unstructured text categorization.