Latent Dirichlet Allocation based multilevel classification
Sunil Bhutada, V. V. S. S. S. Balaram, Vishnu Vardhan Bulusu · 2014
Information processing and knowledge extraction are the two key factors for mining technique. Many models were proposed and implemented successfully on the available information over internet. Automatic Categorization is a machine learning approach which is important for the information processing. In this paper an attempt is made to propose a multilevel classification model using Latent Dirichlet Allocation (LDA) approach. Though the existence of Latent Dirichlet Allocation (LDA) is observed in the literature but a modified model for multilevel classification is presented which is independent of any language. In order to achieve such model many existing proposals were considered like PLSI, which uses the Exceptional Maximization (EM) method only to train the latent classes. The iterative process of Latent Dirichlet Allocation (LDA), which yields the multilevel classification of the corpus. Topic Modeling is used to discover the hidden things that pervade the collection to annotate the documents according to the new topics.