Parameter tuning in updating the sentiment polarity of objective words in SentiWordNet

The past few years has seen a rise of social media through which people provide their opinions regarding various products and general issues. The availability of this valuable data from which the basic sentiment of the people can be extracted has led to a lot of research in the area of sentiment analysis. The sentiment lexicon, SentiWordNet can be used to perform sentiment analysis. Unfortunately a majority of the words in SentiWordNet are objective. These objective words are useless to the process of sentiment analysis. A new approach was proposed by Chihli et al with an aim of providing positive or negative polarity to the objective words. The aim of our work is to further improve upon this approach. Two approaches were tried by us. In the first approach Word Sense Disambiguation was used to find the best sense of the word in SentiWordNet while performing Sentiment Analysis. In the second approach two threshold parameters in the original approach were fine tuned to get the combination of parameters that gave the best accuracy.