Bengali Functional Sentence Classification through Machine Learning Approach

In the early time, very few studies were accomplished in Bengali functional sentences. However, the study on Bengali has incredibly increased for its structural diversity. Inspired by those studies, Functional sentence classification in Bengali language was completed including machine learning approaches to classify the sentences. Three types of Bengali functional sentences such as Assertive, Interrogative and Exclamatory have been considered for the research. So the leading purpose of the study is to classify the sentence and find out the best algorithm with comparing accuracy rate. Data have been collected, categorized and processed the dataset properly to avoid the conflict. Some popular machine learning algorithms such as Naive Bayes (NB), Decision Tree Classifier (DT), SVM, KNN, Random Forest (RF), and XGB Classifier have been implemented to compare accuracy rates. Parameters such as Precision, Recall, F1-Score, Support and Confusion matrix have been calculated for the comparison. The comparison demonstrated that performance of the Random Forest, SVC, and XGB Classifier is better than Naive Bayes and Decision Tree Classifier. Remarkable issue is that the Random Forest algorithm provided the highest performance value with an accuracy of 75.38% which is average performance for such a dataset.

I. Introduction

Language is the best way to express someone's sense where a sentence is the textual unit of language. Actually, a sentence is a set of words that in principle tells a complete thought. From the beginning to now, in this technological world, most of the work, all research works, journals or devices are done following English language. Extensive research is currently being done on other languages than English for the development of computer languages. In the present world, Bengali is the fifth most-spoken native language. Moreover the language is the sixth most spoken by total number of speakers. Researchers are currently conducting various researches on different languages in order to connect languages with Machine Learning and Artificial Intelligence. Accordingly, for the structural special features of the Bengali language and to employ the mother tongue in technology, a lot of work is being done in Bangladesh recently. Stimulated by this, the work has started on the classification of Bengali sentences in this paper. Sentence