Semantic latent dirichlet allocation for automatic topic extraction
Sunil Bhutada, V. V. S. S. S. Balaram, Vishnu Vardhan Bulusu · Journal of Information and Optimization Sciences · 2016
With the inception and uncontrollable growth of digital documents, Automatic Topic Extraction is found to be an active research topic. In order to handle the processing of documents, variety of algorithms was presented in literature for topic extraction using distribution Modeling and classification approaches. Among different modeling methods of topic extraction, Latent Dirichlet Allocation (LDA) is one of the important algorithms for Topic Identification. Even though LDA is popular technique for topic identification, it turns to difficulty in determining the model parameters and also, suffers with finding the degree of similarity and semantic handling. In order to handle these challenges, a new method is proposed for automatic topic extraction. Accordingly, this method called, Semantic Latent Dirichlet Allocation (SLDA) is proposed by extending LDA in a semantic way. A new mathematical computation is included where the model parameters are estimated using new membership degree in Semantic Latent Dirichlet Process along with semantic similarity measure. The experimentation is carried out with two different databases and it observed that SLDA outperformed by showing better when compared with existing LDA in Jaccord Coefficient.