Framework for Analyzing and Visualizing Topics and Sentiments on Social Media: the Case of MH 17 Tweets

This paper develops a framework for sentiment analysis and clustering of Russian and Finnish-language Twitter posts (tweets) that enables the comparison of different language speakers' reactions to the same brand, product, or event. The framework can be further developed to analyze tweets in more than two languages or from more than two regions. Namely, we develop a tweet stream clustering algorithm, which calculates sentiment, keywords, topic tags and growth rate of the harvested tweets. The system utilizes a translation matrix for document vectors in Finnish and Russian to enable comparison of Russian and Finnish-language tweet clusters in a cluster visualization view (sentiment, topics, keywords, growth rate of clusters). A map view also allows following the density of Twitter activity in a given geographical location. To illustrate the framework, we examine the Russian and Finnish language tweets about the Malaysia Airlines Flight 17 (MH17) crash on July 17, 2014.