A Novel Graph Based Framework to build Multi Label Text Classifier
Maya Ingle, Parag A. Kulkarni · 2012
Text document is multifaceted object and associated with many properties such as multi labeledness. Under this a single text document can inherently belongs to more than one category simultaneously. Traditional single label and multi class text class ification paradigms cannot efficiently classify such multifaceted text corpus. Through our paper we are proposing a graph based frame work for Multi Label Text Classification paradigm. Representing text documents in the form of graph vertices rather than the vec tor representation like Bag of Words allows pre-computing and storing of necessary information. It also models the relationship between text documents and class labels. We are using semi supervised learning technique in our proposed approach for effectively utilizing labeled and unlabeled data for classification .Our proposed approach promises better classification accuracy and handling of complexity. Our propos ed framework is elaborated on the basis of standard dataset such as Enron, Slashdot, Bibtex and Reuters.