Investigating Temporal and Spatial Trends of Brand Images Using Twitter Opinion Mining

Abstract

The growing popularity of social network services has led to many studies of various phenomena in this area. However, most of this research has been conducted using English language data, and relatively little has considered Korean. In this paper, we demonstrate a systematic analysis framework using Korean Twitter data to mine temporal and spatial trends of brand images. A publicly available Korean morpheme analyzer is used to analyze Korean tweets grammatically, and we construct Korean polarity dictionaries containing a noun, adjective, verb, and/or root to automatically analyze the sentiment of each tweet message. Sentiment classification is performed by a support vector machine and multinomial naïve Bayes classifier. In particular, our own feature selection step improves the support vector machine sentiment classification accuracy to 80%. Based on this result, we visualize the temporal and spatial distribution of brand images, and present the temporal changes of brand-related keyword networks. Our analysis enables trends in brand awareness to be systematically traced and evaluated. This allows various other analyses, such as advantages and disadvantages of the brand, and a comparison with its competitors.