A Hybrid Optimized Model for Opinion Mining with Dual Weightage Aspects

Opinion mining, natural language processing technique task is used for extracting information from unstructured data available on the internet. Now a days, users’ shared aspect-based online views have become a reliable source of information for customers making purchasing decisions and business purposes. In this paper, a new aspect/feature based opinion mining approach with four primary components: preprocessing, extraction of aspect sentiment, grouping of aspect term and Aspect Sentiment Classification are discussed. The preprocessing of the user’s reviews is performed by tokenization, lemmatization, stemming and removal of stop words. The extraction of aspect sentiment is carried out in two stages: POS tagging and Rule-based lexicon extraction. Based on the estimated semantic similarity score, the retrieved characteristics are divided into primary and secondary aspects. Both the primary and secondary aspects are then multiplied by the weight function. Furthermore, the suggested Particle with Levy Updated Sparrow Search Algorithm model optimizes the primary and secondary weighting elements to make the categorization more suitable in terms of opinion. The hybrid optimization model is created by combining the conventional Particle Swarm Optimization (PSO) with sparrow search algorithm (SSA). Following that, the retrieved attributes for final classification take place via Neural Network. The opinions of the reviews are characterized as a positive, neutral or negative.