Exploration of sentiment analysis and legitimate artistry for opinion mining

Abstract

Sentiment analysis/opinion mining is a technique that analyzes people’s opinions, evaluations, sentiments, attitudes, appraisals and emotions to entities like products, organizations, services, issues, individuals, topics, events and their attributes. It is a massive problem space. People tend to express their opinions on anything, such as, a product, service, topic, individual, organization, or an event. Here, the term object represents the entity commented on. Certain private states parts that cannot be judged and observed include the following, beliefs, opinion, emotions and sentiments. The above mentioned aspects are usually expressed in documents using certain subjective words that determine the private states with the help of unique dictionaries like the WordNet or SentiWordNet. The feature selection concept is incorporated in the following tasks such as image classification, data mining, cluster analysis, image retrieval, and pattern recognition. This is observed as a data analysis pre-processing strategy; here a subset from the original data features is thus selected for eliminating the noisy/irrelevant/redundant features. This technique essentially helps in minimizing the incurred computational expenses and helps in enhancing the accuracy level of the data analysis procedures. The Semantic features are meant to concentrate on the relationship between the signifiers such as that of the words, phrases, signs and symbols. A special of semantics called as the linguistic semantics is used for understanding human based expression in opinions and blog. A semantic based feature selection strategy has been introduced for establishing the opinion mining tasks. This introduced semantic based feature selection makes use of the SentiWordNet that is observed to be a lexical resource of the WordNet database extracted terms and is therefore used in the research tasks. Feature set is minimized with the help of the introduced semantic based approaches for the purpose of considering the individual predictive ability words and selection features. Experiments were conducted with the help of the Naive Bayes, the FLR and the AdaBoost classifiers and the obtained results were compared for understanding and judging the feature selection methods.