AOQAS: Ontology Based Question Answering System for Agricultural Domain

Abstract

Agriculture is an indispensable sector for human community that has been transformed by technological innovations. The data handling with information extraction is one of the areas that is benefited by the advancements in information technology. The presented research work aims to develop a question answering system (QAS) for improving the information retrieval from the agricultural text documents. The proposed Agriculture domain Ontology based QAS (AOQAS) processes the given agricultural text documents and constructs it to a knowledge representation called ontology. The domain based ontology is created using the Bidirectional Encoder Representations from Transformers model (BERT model) with Regular Expressions (RE) for withdrawing domain terms and the Bidirectional Long Short Term Memory model (BiLSTM) with RE for relationship extraction between the agricultural terms. From the developed ontology, the answers for the input query are extracted and validated using Natural Language Processing (NLP) techniques and the deep learning model. The proposed AOQAS shows an accuracy and recall of 98.47% and 98.26%. The outcomes of AOQAS shows better performance when it is evaluated against the current systems.