Machine Learning Approaches for Bengali Automated Question Detection System

Question detection is the initial tasks of question answering system. Question detection, the first step of QA system has been done applying SVM, Logistic Regression, K-Neighbor classifier, Multilayer Perceptron and LSTM. We achieved best performance for SVM with linear kernel trick. For 2000 feature words we got 1.4% error rate which is best among the approaches we used for Bengali language. For LSTM algorithm we got 3.22% error rate. LSTM looks promising with large and better dataset.