Deep ensemble approach for question answer system

Abstract

Researches on question answering systems has been attracting significant research attention in recent years with the explosive data growth and breakthroughs in machine learning paradigm. Answer selection in question answering segment is always considered as a challenging task in natural language processing domain. The major difficulty detected here is that it not only needs the consideration of semantic matching between question answer pairs but also requires a serious modeling of contextual factors. The system aims to use deep learning technique to generate the expected answer. Sequential ensemble approach is deployed in the proposed model, where it categorically boosts the prediction of LSTM and memory network to increase the system accuracy. The proposed model shows a 50% increase in accuracy when compared to individual systems with a few number of epochs. The proposed system reduces the training time and boosts the system-level accuracy.