Non-parametric Method of Topic Identification Using Granularity Concept and Graph-Based Modeling

This paper aims to classify the large unstructured documents into different topics without involving huge computational power and a priori knowledge. The concept of granularity is employed here for extracting contextual information from the documents by generating granules of words (GoWs), hierarchically. In a novel computationally efficient w ay, the proposed Granularity Based Word Grouping (GBWG) algorithm exploits the co-occurrence between the words of different granules to group the contextually related words, representing different topics. However, the topics are coarsely generated, containing common words. A Graph-based method has been proposed to precisely identify the topics using the Louvain graph clustering algorithm. The algorithm clusters the words using mutual information as an association measure between the words of each granule. Each cluster containing unique words, represents different topics. Finally, a test document is classified into a particular topic, corresponding to the cluster having the maximum probability of belongings of the words in the test sequence. The Performance of the proposed method has been compared with other unsupervised, semi-supervised and supervised topic modeling algorithms and we demonstrate that the result is comparable or better than the state of the art algorithms.