Suitability of Naïve Bayesian Methods for Paragraph Level Text Classification in the Kannada Language using Dimensionality Reduction Technique

The amount of data present online is growing very rapidly, hence a need for organizing and categorizing data has become an obvious need. The Information Retrieval (IR) techniques act as an aid in assisting users in obtaining relevant information. IR in the Indian context is very relevant as there are several blogs, news publications in Indian languages present online. This work looks at the suitability of Naïve Bayesian methods for paragraph level text classification in the Kannada language. The Naïve Bayesian methods are the most primitive algorithms for Text Categorization tasks. We apply dimensionality reduction technique using Minimum term frequency, stop word identification and elimination methods for achieving the task. It is evident that Naïve Bayesian Multinomial model outperforms simple Naïve Bayesian approach in paragraph classification tasks.