Sentiment analysis for design of computing with words based recommender

Abstract

Sentiment analysis is a remarkable machine learning technique that accepts important words from a given text. It gives the output in the form of a sentiment score of each of the word and the overall text along with orientation of both (keywords and text) as being positive/ neutral/ negative. Social media has become a new platform for discussions in recent years due to growth in its reach. Sentiment analysis has proved to be quite useful in determining the collective response of people on any issue by analysing their opinions which are generally in the form of written texts. However, human beings do not understand numbers but language. So, we use the mathematical technique of perceptual computing to provide a mapping between numeric and linguistic data to generate recommendations. It is based on Zadeh's computing with words (CWW). However, to generate recommendations from perceptual computing, we need problem specific linguistic terms and their associated interval values. Obtaining interval values may not be possible in scenarios where the number of subjects available for providing feedback is very less. So, here we propose a new approach that processes the linguistic information using perceptual computing based on the intervals extracted using the sentiment score using a sentiment analysis tool.