Indian Journal of Science and Technology

Abstract

Objectives: To find the use of emoticons as indicators in an unsupervised sentiment analysis system. Methods/Analysis: Sentiment analysis is the concept of extracting opinions and assigning different sentiment to the collected opinions. A sentiment analysis experiment using synthetic data set of text with simulated emoticons/reactions has been proposed and evaluated the ability to cluster sentiment with two clustering algorithms, namely k-means and Agglomerative Clustering. Findings: It was found that emoticons are powerful indicators, and made a discussion of a system where they are used to augment a text-only based system. Novelty / Improvement: The algorithms produce better sentiment clustering with emoticon data and their performance has been differentiated by using clustering performance evaluation method. 

Keywords: Agglomerative, Clustering, Emoticons, k-means, Sentiment Analysis, ARI