Polarity Shift Detection and Handling in Polarity Classification: A Survey

Abstract

User-generated data over the internet is highly opinionated. Sentiment analysis is a task to extract opinions without losing any crucial information. Polarity classification is most researched from the last two decades under sentiment analysis. However, Polarity classification is a tedious job in natural language processing. There are certain circumstances where the polarity of the phrase differs from the polarity of words into it. This inconsistency is due to some trigger words that make the word polarity to get reversed. One of the main reasons for the performance degrade of the classification algorithm is the polarity shift. Polarity shift detection and handling approaches are applied to detect and handle polarity shifts, which can enhance the accuracy and effectiveness in classification. This paper gives a better understanding of the issue of polarity shift and surveys various methods of polarity shift detection and handling.