Title: Need for Hybrid Lexicon Based Context Aware Sentiment Analysis for Handling Uncertainty—An Experimental Study
Abstract: In this modern era, the rapid improvement of internet technologies makes the user comfortable in generating the data in an easier way. To analyse the user-generated data for classifying sentiment polarity into one of the three categories namely positive, negative and neutral, sentiment analysis is required. Sentiment analysis is the computational study of user opinions, moods, sentiments and other subjective elements of the text. Sentiment analysis can be implemented using lexicon based approaches and machine learning approaches (Bnadhane et al. in Procedia Comput Sci 45: 808-814 (2015), [1]). Sentiment lexicon can be used to maintain terms and their respective sentiment values. But, the existing sentiment lexicons cannot handle improved internet slang data and missing data. Hybrid lexicon can be generated by combining sentiment lexicon with domain-specific sentiment-bearing terms. In this work, we analyse and find the need for improving sentiment classification with derived knowledge from domain-specific contextual analysis and domain adopted lexicons. Finally, the analysis shows the need for the proposed sentiment classification to handle missing data.