Ontology-Based Natural Language Processing for Sentimental Knowledge Analysis Using Deep Learning Architectures

Abstract

When tested with popular datasets, sentiment categorization using deep learning (DL) algorithms will produce positive results. Building a corpus on novel themes to train machine learning methods in sentiment classification with high assurance, however, will be difficult. This study proposes a way for representing efficient features of a dataset into a word embedding layer of DL methods in sentiment classification known as KPRO (knowledge processing and representation based on ontology), a procedure to embed knowledge in the ontology of opinion datasets. This research proposes novel methods in ontology-based natural language processing utilizing feature extraction as well as classification by a DL technique. Here, input text has been taken as web ontology based text and is processed for word embedding. Then the feature mapping is carried out for this processed text using least square mapping in which the sentiment-based text has been mapped for feature extraction. The feature extraction is carried out using a Markov model based auto-feature encoder (MarMod_AuFeaEnCod). Extracted features are classified by utilizing hierarchical convolutional attention networks. Based on this classified output, the sentiment of the text obtained from web data has been analyzed. Results are carried out for Twitter and Facebook ontology-based sentimental analysis datasets in terms of accuracy, precision, recall, F-1 score, RMSE, and loss curve analysis. For the Twitter dataset, the proposed MarMod_AuFeaEnCod_HCAN attains an accuracy of 98%, precision of 95%, recall of 93%, F-1 score of 91%, RMSE of 88%, and loss curve of 70.2%. For Facebook, ontology web dataset analysis is also carried out with the same parameters in which the proposed MarMod_AuFeaEnCod_HCAN acquires accuracy of 96%, precision of 92%, recall of 94%, F-1 score of 91%, RMSE of 77%, and loss curve of 68.2%.