Myanmar Sports Sub-News Categorization Using Chi-Square, Bag of Words, and Cosine Similarity Combining with Smooth TF-IDF

In recent years, text news, unstructured data, are rapidly growing over the Internet. To automatically categorize and organize text news, text classification is used. It is applied in topic labeling, spam detection, and sentiment analysis. This paper proposes a hybrid method for automatic Myanmar sports sub-news categorization. Myanmar sports news has been categorized into 19 sports sub-news categories by combining Chi-Square and bag of words (BOW) based system and the term frequency-inverse document frequency (TF-IDF) based system. The classification method used for the systems is cosine similarity. The experimental results showed that the hybrid approach achieved better accuracy (98.85%) than the separated approaches (Chi-Square and BOW based system 96.18% and TF-IDF based system 93.13%) without hybridization.