Twitter Event Detection, Analysis, and Summarization

Data analysis in social media is a broad and well-addressed research topic, but the characteristics and sheer volume of Twitter messages with high amounts of noise in them make it a difficult task for Twitter. Tweets reporting real-lifhe TWItter event Summarizer and Trend detector (TWIST) system that attempts to tackle these challenges by combining wavelet and text analysis. The system detects and analyzes real-life events reported in Twitter. For providing high-quality summaries with clean and meaningful content, TWIST analyzes external sources for detected events, which it accomplishes by retrieving links and extracting the main tweets describing those events. Then, all detected events are analyzed by their sentiment distribution over a world map, and visualized in the resolution of countries. This feature enables the TWIST user to see whether, and if so, how the geolocation of Twitter users affects their opinions, and how the sentiments and opinions regarding the same event can be different over different countries. Then, by analyzing wavelet form and user activity, a prediction can be made for an event to be highly active. In our approach, we utilize unsupervised learning for most stages, except sentiment analysis. Our system does not rely on external ontologies, but incorporates external sources, automatically retrieved, for summarizing events. Also, TWIST uses geolocation of event-related tweets to visualize their sentiments on a map. Finally, we use text and hashtag analysis in order to predict high-activity events before they are fully developed. Because our approach requires basic text preprocessing only, it can therefore be easily adapted to multiple languages.e events are usually overwhelmed by a flood of meaningless information. This chapter describes t.