An Improved Approach for Topic Ontology Based Categorization of Blogs Using Support Vector Machine

Abstract

Problem statement: Information search, collection and categorization from the blogosphere are still one of the important issues to be resolved. Mainly, the blogs assist the variety of interesting and useful information. Because of its increasing growth, blogs can not be categorized effectively. Therefore it is difficult to find relevant topics from the blogs. Hence blogs need to be categorized topically to make easy for readers. Approach: Blog contents are associated with a set of predefined topic ontology keywords. This study proposes categorization of blogs to facilitate easy identification of user expected topic from the massive collection of blogs. Tags, page contents were collected as inputs from the blogs and the blogs were categorized using Support Vector Machine (SVM) algorithm. Most frequent occurrences of topic ontological keywords are used to train the classifier. This approach has effectively improved blog categorization process using SVM. Results: The performance was evaluated for precision and recall for blog categorization based on topic ontology using SVM with Naive Bayes algorithm. It was proved that topic ontology assisted SVM improves the classification accuracy than Naïve Bayes algorithm. Conclusion: This study has effectively improved the classification of blogs based on topic ontology assisted SVM. Experiments showed the effectiveness of the blog categorization.