SentimentState: Exploring Sentiment Analysis on Twitter

Contributors

Taurean Scantlebury, Christopher Collins

Abstract

Twitter feeds are a potential source of useful information regarding the state of mind of persons who are the subject of legal or medical assessment. These may include persons suspected of committing crimes or patients that arrive at a hospital for a mental health emergency, for example, attempted suicide. Messages called “tweets” can expose the state of mind of a Twitter user.  Analysts are challenged with creating reports of the online presence of users quickly and efficiently. We present a web-based visualization tool called SentimentState that performs sentiment analysis on tweets from a user’s Twitter account.

SentimentState analyses tweets based on ten emotions (positive, negative, anger, anticipation, disgust, fear, joy, sadness, surprise and trust) and creates an interactive time-line graph of the emotional state of the user. It uses a collection of emotion 24,200 word-sense pairs collected from the National Research Council of Canada (NRC). We anticipate that this interactive visualization can have applications throughout, and even beyond, legal and medical assessments, and will provide analysts with timely and relevant information regarding the mood state of clients, patients and other persons under assessment.

Online Demo

Try the visualization yourself!

Resources

GitHub Repository

Publications

• T. Scantlebury, “SentimentState: Exploring Sentiment Analysis on Twitter,” Bachelors Thesis, 2014.
  [Bibtex]

  @BachelorsThesis{sca2014a,
    author =    {Taurean Scantlebury},
    title =    {SentimentState: Exploring Sentiment Analysis on Twitter},
    school = {University of Ontario Institute of Technology},
    year =   2014,
  }

Acknowledgements

Thanks to Saif Mohammed for providing the NRC Emotion Lexicon for this project.