Comparing and combining sentiment analysis methods

A new method that combines existing approaches, providing the best coverage results and competitive agreement is developed and a free Web service called iFeel is presented, which provides an open API for accessing and comparing results across different sentiment methods for a given text.

Abstract

Several messages express opinions about events, products, and services,\npolitical views or even their author's emotional state and mood. Sentiment\nanalysis has been used in several applications including analysis of the\nrepercussions of events in social networks, analysis of opinions about products\nand services, and simply to better understand aspects of social communication\nin Online Social Networks (OSNs). There are multiple methods for measuring\nsentiments, including lexical-based approaches and supervised machine learning\nmethods. Despite the wide use and popularity of some methods, it is unclear\nwhich method is better for identifying the polarity (i.e., positive or\nnegative) of a message as the current literature does not provide a method of\ncomparison among existing methods. Such a comparison is crucial for\nunderstanding the potential limitations, advantages, and disadvantages of\npopular methods in analyzing the content of OSNs messages. Our study aims at\nfilling this gap by presenting comparisons of eight popular sentiment analysis\nmethods in terms of coverage (i.e., the fraction of messages whose sentiment is\nidentified) and agreement (i.e., the fraction of identified sentiments that are\nin tune with ground truth). We develop a new method that combines existing\napproaches, providing the best coverage results and competitive agreement. We\nalso present a free Web service called iFeel, which provides an open API for\naccessing and comparing results across different sentiment methods for a given\ntext.\n