Telugu Lyrics Based Classification By Using Naive Bayes
Abdul Aziz, Ganga Raju Ravula, Mobeen Taj Shaik, Sravanthi Potharaju · 2020
The main objective of this research is to predict the song category by using a Naive Bayes classifier. This type of research was previously done for Telugu songs using both audio and lyrical features. In that research they have addressed the lyrics as a whole, beginning, and the ending of the song. When they have given the whole song to model the accuracies are low, while they used both Naive Bayes and SVM for classification. Now this paper describing how Naive Bayes itself only holds the classification results more accurate with the whole song given as input to the model. This paper mainly holds the concepts of data pre-processing, feature extraction, and text classification to evaluate the model. The dataset consists of lyrics collected from four different genres, such as Melody, Sad, Rainy, and Pelli (marriage). This proposed method performs classification and calculates accuracies for the given dataset. The final accuracy obtained for this model is 92.3% using the Naive Bayes classifier.