Senti-NSetPSO: large-sized document-level sentiment analysis using Neutrosophic Set and particle swarm optimization

In the last decade, opinion mining has been explored by using various machine learning methods. In the literature, document-level sentiment analysis has been majorly dealt with short-sized text only. For large-sized text, document-level sentiment analysis has never been dealt. In this paper, a hybrid framework named as "Senti-NSetPSO'' is proposed to analyse large-sized text. Senti-NSetPSO comprises of two classifiers: binary and ternary based on hybridization of particle swarm optimization (PSO) with Neutrosophic Set. This method is suitable to classify large-sized text having more than 25 kb of size. Swarm size generated from large text can give a suitable measurement for implementation of PSO convergence. The proposed approach is trained and tested for large-sized text collected from Blitzer, aclIMDb, Polarity and Subjective Dataset. The proposed method establishes a co-relation between sentiment analysis and Neutrosophic Set. On Blitzer, aclIMDb and Polarity dataset, the model acquires satisfactory accuracy by ternary classifier. The accuracy of ternary classifier of the proposed framework shows significant improvement than review paper classifier present in the literature.