Exploiting Distance Graph and Hidden Topic Models for Multi-label Text Classification

Abstract

Hidden topic models, the method to automatically detect the topics which are (hidden in a text) represented by words, have been successfully in many text mining tasks including text classification. They help to get the semantics of text by abstracting the words in text into topics. Another new method for text representation is distance graph model, which has the ability of preserving the local order of words in text, thus, enhancing the text semantics. This paper proposes a method to combine both hidden topic and distance graph models for opinion mining in hotel review domain using multi-label classification approach. Experiments show the efficiency of the proposed model provides a better performance of 4% than that of the baseline.