Multiple Choice Question Answer System using Ensemble Deep Neural Network

Abstract

Question answer (QA) system is closely related to NLP and IR tasks. An automated QA system should understand the semantics of question and derive answers relevant to it. In case of MCQ system this tasks becomes difficult as the model needs to understand the semantics and select an answer from a given choice. In this paper we propose a ensemble approach to predict answers to Multiple choice question using LSTM model, hybrid LSTM –Convolution NN model and Multilayer Perception (MLP) model. Firstly, by using LSTM and hybrid LSTM-CNN models are trained parallel. Multilayer Perception is used to predict option to training dataset separately. The 8thGr-NDMC datasets are selected for model evaluation and comparison. The 8th GR-NDMC is used for experimentation purpose. The observed results demonstrate that the proposed approach performs better than some other single forecasting models.