Automated Categorization and Analysis of Kannada News Content Using Transformers

Online news consumption has increased dramatically in the modern era, making effective classification methods for news content necessary. The process of systematically classifying news items so that readers can get pertinent information without needless effort is known as news categorisation. An efficient classification system is necessary to enhance accessibility and user experience because many sources publish large volumes of news stories every day. Using a dataset of 700 articles in seven different categories, this study concentrates on automated news category classification. We created our own dataset for particular categories, such as politics, crime, business, and technology, in order to overcome the problem of inconsistent data. We developed a supervised machine learning model to classify articles into business, entertainment, politics, sports, technology, lifestyle, and crime using the pre-trained XLMRoBERTa model. By screening articles according to user interests, this method optimises user engagement and allows for personalised content distribution. With an F1-score of 60.06% and a classification accuracy of 65.71%, our model proved effective in managing the multilingual and context-sensitive data found in online news.