Twitter Bullying Detection

Data mining of Social networks is a new but interesting field within Data Mining. We leverage the power of sentiment analysis to detect bullying instances in Twitter. We are interested in understanding bullying in social networks, especially in Twitter. To best of our understanding, there is no previous work on using sentiment analysis to detect bullying instances. Our training data set consists of Twitter messages containing commonly used terms of abuse, which are considered noisy labels. These data are publicly available and can be easily retrieved by directly accessing the Twitter streaming API. For the classification of Twitter messages, also known as tweets, we use the Naïve Bayes classifier. It‟s accuracy was close to 70% when trained with “commonly terms of abuse” data. The main contribution of this paper is the idea of using sentiment analysis to detect bullying instances.