Prediction Of Emotions In Kannada Sentence With Homonyms

The sentimental analysis especially with homonym words in vernacular language is a challenging process. Even though Kannada experts have studied the classification of emotions, there is still misunderstanding when words have diverse meanings. This research focuses on the categorization of Emotions and No Emotions in Kannada sentences, leveraging machine learning techniques particularly Support Vector Machines (SVM) and Naive Bayes. The work uses preprocessing approaches to solve linguistic oddities with a dataset of 271 (160 Emotion and 110 No Emotion) hand labelled Kannada sentences. Feature extraction involves methods such as Bag-of-words and TF-IDF. Emotion recognition models are created using the SVM and Naive Bayes algorithm. Naive Bayes demonstrated the maximum accuracy at 70.90% while SVM performed with an accuracy of 56.36%.