Automatic Detection of Opinion Polarity and Their Strengths in Newspaper Editorials

Abstract

In this paper, we discuss the development of our methods in performing analysis on the newspaper editorial. We discover that analyzing editorial texts for opinion polarity and strength is a daunting task. It poses several challenges – implicitness of opinions, unpredictable situation with opinion polarity and strength, variance in the language expressions and so on. We evaluate the resource based approach in detecting opinion polarity and strength thereby measuring the effectiveness of the different lexicons and rules provide current results before the last line. The results show that the current accuracy are quite low and this approach requires further efforts to achieve better scores.