Event Based Sentiment Analysis of Twitter Data

Everyday large volumes of data are produced. Millions of users share and dissipate most up-to-date information on twitter. Traditional text mining suffers severely from short and noisy nature of tweets. Event detection from twitter data has many new challenges when compared to event detection from traditional media. Noisy nature and limited length are the challenges imposed by twitter data. Event detection performance on twitter is negatively affected by nature of tweets. This paper proposes SegAnalysis framework to tackle these challenges. It performs tweet segmentation, event detection and sentiment analysis. Tweet segmentation is performed in a batch mode using POS (part of speech) tagger on recent online tweets fetched by the user. Segmentation of a tweet preserves the named entities and its stickiness score is calculated. Naïve Bayes classification and online clustering detect events. These events improve situational awareness and decision support. Sentiment analysis categorizes tweets as positive, negative and neutral depending on sentiment score of a tweet. SegAnalysis framework can be extended to deal with events belonging to multiple clusters.