Urdu news categorization using machine learning approaches

Rapid technological developments have changed the way of presenting news content in media. The last three decades have witnessed the emergence of digital media platforms that offer a broad variety of news content ranging from business to, economics, from Science&Technology to sports. Due to massive amount of data produced, the need for automatically categorizing them has arisen. Motivated by this, we have addressed the problem of text categorization of Urdu news by using four machine learning approaches namely (Naïve Bayes, Support Vector Machines, Logistic Regression and Decision tree). In the first step we have collected the data which contains 4000 news belonging to four different categories (Sports, Business-&-Economics, Science-&- Technology and Entertainment). The data is in the raw format so before setting up the machine learning model, we applied pre-processing techniques like tokenization, removing the stop words and lemmatization. Then, the features are extracted Bag of Words and Term Frequency-Inverse Document Frequency methods. In the last step, we evaluate the performance of machine learning algorithms utilizing accuracy, precision, recall, and F1-score metrics. In future work, deep learning models can be can be utilized provided that more data is collected and a system can be developed which performs real-time categorization.