Development of a Closed Domain Question Answering System based on Deep Learning

Abstract

Close Domain Question Answering systems generates answers to questions asked for a specific field of interest like military, health and medicine, economics, education and finance. The process involves information retrieval tasks that automatically extracts answers to the questions asked by humans in natural language using either a pre-structured database or a collection of natural language documents. Deep learning approach was employed in this research to develop a closed Domain Question Answering System for a Nigerian Tertiary institution. The corpus used in training the system was pre-processed using Pandas Library and an end to end comprehension CdQA based on Bidirectional Encoder Representations from Transformers (BERT) as used in pre-training the language representations. The GUI was designed using python tool PyQt5 and the performance of the system was evaluated using Precision, Recall and F-Measure metrics. Results from evaluation with 50 structured questions
gave precision score of 88%, Recall rate of 76% and F-Measure rate of 0.82. The results from the evaluation metrics show that the system is efficient and gives a higher accuracy and precise answers to structured questions.