Comparison of web services for sentiment analysis in social networking sites

Abstract

With various type of web services available, it is hard to identify and compare which of the free access web services work best in analysing sentiment of extremist content in social networking sites. For that purpose, a generic approach by working with API of web service using PHP programming language is used to test each dataset that was extracted based on the keyword 'extremism'. Data from both Twitter and Facebook has been used as these two are the most powerful platforms for expressing one's feeling. The comparison for web service is done based on the analysis of its accuracy, precision, recall and f-measures in obtaining the lowest score of mean square error (MSE). Four sentiment analysis web services are used which are Sentiment Analyzer, Aylien, ParallelDots, and MonkeyLearn. From the comparison, MonkeyLearn obtained the best final results among all web services with the lowest MSE score of 14%. For the benefit of other researchers, the finding of this will reveal the suitable web service for analysing sentiment issues as critical as extremism.