Sentiment analysis of Arabic comparative opinions

The rapid development of social media platforms allowed opinion mining research to increase significantly. Opinion mining/sentiment analysis is the process of extracting subjective information from opinions that imply single sentiment. Comparative opinion mining is a sub-field of opinion mining that deals with multi-sentiment opinions. Such opinions are expressed by comparing several entities to each other. The sentiment of a comparative relation is recognized by identifying the relation’s direction and thus the preferred entity. This paper proposes an unsupervised sentiment analysis technique for Arabic comparative opinions to identify the preferred entity. The proposed technique considers three main elements when analyzing comparative opinions: The type of comparative keywords, the existence of features in the opinion, and the entities’ position to the comparative keyword. Five main categories are proposed for classifying comparative keywords, which facilitates the analysis of each comparative sentence. The proposed technique limits the need for human interference to the initial steps of preparing the lexicons, collecting and categorizing comparative keywords. Furthermore, the proposed technique handles opinions that do not contain features at all. The results are very promising with a total average of 96.5% f-measure of correctly identified sentiment.