Latent Topic-based Graph Construction for Text Classification

Abstract

This paper aims to raise the accuracy of multi-class text classification by means of graph-based semi-supervised learning (GBSSL). It is essential to construct a proper graph expressing the relation among nodes in GBSSL. We propose a method to construct a similarity graph by employing both surface information and latent information to express similarity between nodes. Experimenting on Reuters-21578 corpus, we have confirmed that our proposed method works well for raising the accuracy of GBSSL in multi-class text classification task.