Hybrid Intelligent Technique for Text Categorization

Text categorization is the task in which documents are classified into one or more of predefined categories based on their contents. This paper shows that the proposed system consists of three main steps: document representation, classifier construction and performance evaluation. In the first step, a set of pre-classified documents is provided. Input documents are initially pre-processed in order to be split into features and eliminate non-informative features. The remaining features are next weighted based on the frequency of each feature in that document and standardized by reducing a feature to its root using the stemming process. Due to the large number of features even after the non-informative features removal and the stemming process, the proposed system applies specific thresholds to extract distinct features which represent the input document. In the second step, the text categorization model (classifier) is built by learning the distinct features which represent all the pre-classified documents, this process can be achieved by using one of the supervised classification techniques that is called the rough set theory. The model uses a pair of precise concepts from the above theory that are called lower and upper approximations to classify any test document into one or more of main categories and sub-categories. In the final step, the performance of the proposed system is evaluated. It has achieved good results up to 96%, when applied to a number of test documents for each sub-category of main categories.