Sentiment based emotion classification in unstructured textual data using dual stage deep model

Classification of sentiments is an essential task in Natural Language Processing (NLP) domain. The powerful sentiment classification helps determine user opinions in product reviews or social networks. However, comprehending hidden opinions, sentiments, and emotions in emails, tweets, reviews, and comments is a challenge and equally crucial for social media monitoring, brand monitoring, customer services, and market research. Also, unstructured data in social media remains a major issue, and a proficient technique to deal with this issue remains a research gap. Therefore, this work presents an automated sentiment polarity and emotion classification in unstructured textual data using dual stage deep learning framework. Initially, pre-processing is performed to remove the noises and promote the quality of input data using stop words removal, Parts-Of-Speech (POS) tagging and duplicates removal. Then, the most discriminative features are extracted in the feature extraction stage, and the optimal set of features is selected to minimize the large feature dimensionality. Finally, the selected features are provided to the Dual-stage Deep Model to classify sentiments and emotions. The proposed classification stage classifies the sentiment and emotions from the given input data. The proposed work used three datasets for simulation analysis, and each dataset’s performance is determined. Using Twitter Sentiment Dataset, the proposed model obtains an accuracy of 99.80%, F1-measure of 99.667%, specificity of 99.85% and kappa value of 99.52%, IMDB Movie Reviews attains an accuracy of 99.75%, F1-measure of 99.47%, specificity of 99.75% and kappa value of 98.99% and Yelp Reviews Dataset attains accuracy of 99.83%, F1-measure of 99.6%, specificity of 99.83% and kappa value of 99.32%. The obtained results reveal the effectiveness of a proposed study.