A Hybrid Optimized Deep Learning Framework to Enhance Question Answering System

One of the challenging tasks in big data machine learning is the Question-Answering (QA) system. The QA system datasets have several question types: multiple-choice questions, yes or no queries, Wh questions, short questions, factoid questions, etc. Henceforth, training the data and classification of questions and answers is a difficult task. To address these issues, the current research work has focused on constructing a novel Recurrent Hybrid Ant Colony and African Buffalo Model (RHAC-ABM) for the QA classification and answer selection process. Initially, the specific QA dataset was trained to the system, then the trained datasets were tokenized, and training flaws are removed systematically. Moreover, the proposed model was designed by upgrading the hybrid Ant and African buffalo fitness to a dense layer. Also, the hybrid function in the recurrent classification layer can afford a high accuracy rate for query specification and answer selection process. Subsequently, the proficient measure of the designed approach was validated with other related existing models by comparing the chief metrics. Henceforth, the developed RHAC-ABM has gained 99.6% accuracy, F-measure, 99.51%, recall 98.5%, precision 98.5% and low error rate as 1.4% for question classification and answer selection process. Moreover, the achieved results are quite better than other models, it has proved the robustness of the proposed system.