Knowledge based imputation technique for non-opinionated sentences in sentiment analysis

Online reviews play an important role in sales and production of a product. Many researchers apply sentiment analysis on online reviews of any product to classify opinions of users’ views and ideas. Because of the overwhelming unclassified opinions of twitter data, an opinion mining system is required to classify reviews and extract useful knowledge out of them. However, many proposed sentiment classification algorithms consider only the opinionated sentences and omit the sentences without any sentiment words though the aspect of the particular product is present in the reviews. But these non-opinionated sentences have great impact in calculating the accuracy of the aspect-based sentiment score. In this paper, a novel knowledge based imputation technique (KBIT) is proposed to handle these non-opinionated sentences by imputing missing sentiments to improve the accuracy of the aspect based sentiment analysis.