Social media and real world events

If social media is a mirror of what is happening in the real world, then it should be possible to use social media analysis to understand what is happening with real world events. This can help track natural disasters, or even provide statistical models that can help to predict future events.

Twitter has provided a particularly rich seam of research in this area, as most Twitter users have a public twitter stream, meaning that all of their tweets are visible. These tweets can be downloaded on-masse and analysed to see what is happening around a particular keyword, person or event.

The sheer number of mentions can be an important indicator. During the German Federal Election Tumasjan et al (2010) found that the number of tweets correlated with election results, and co-mentions of different parties correlated with real-world ties or coalitions. Tsou and Yang (2012) saw similar results in the 2012 US presidential primaries, with the number of tweets mentioning each candidate in each state correlating strongly with the election result.

These correlations have led some to claim that social media sites like Twitter can be used predictively. Inevitably this has been used to try and predict stock market behavior (for example, Mao et al (2013) used Twitter spikes to build a predictive algorithm that outperformed the market).

The fact that Tweets can also be geotagged (marked with information that tells you where the tweet was written) has also meant that researchers have looked at how it can be analysed in order to get new information about places and locations. For example, Lee at al. (2013) used a content analysis of geotagged Tweets in order to characterise urban areas (automatically finding shopping, office or entertainment areas), effectively crowdsourcing what would normally be an expensive and time-consuming activity.

Others have turned to Twitter to understand how people behave in times of crisis. De Longueville et al (2009) analysed Twitter use during the 2009 forest fires near Marseille, France. They looked at how Twitter was used, and whether it could be used to build a location-picture of the fire developing. Noting that distinguishing between primary and secondary information (eye-witness vs. rumour) was challenging.

Nevertheless this kind of analysis has been used successfully to detect earthquakes in Japan with around 96% accuracy (Sakaki et al. 2010), and also to detect the spread of H1N1 in the US (Signorini et al, 2011). These researchers showed that the social media activity accurately tracks official disease reports, and their system, based solely on social network analysis, would allow disease to be tracked in near real time (as opposed to 1-2 week delay using traditional reporting methods).

Many of these initial studies have been rather simplistic in their methodology, using correlation, or volume to show big patterns, but more nuanced studies are now being undertaken to understand how social media is used during crises. For example, Vieweg et al. (2010) have looked at how Twitter can be used to enhance situational awareness (the big picture) during crisis situations. They conducted two studies (during the Oklahoma Grassfires of April 2009 and the Red River Floods in March and April 2009) and identified different types of information (including warnings, advice, weather updates, road conditions and volunteer information). These kinds of usage models could lead to more sophisticated analysis methods in the future.

How would you feel if one of your social media posts about an event was used in this kind of analysis, are you always a reliable source of data? How would you make judgements about the accuracy of information that other people posted about the event?

Further reading

Optionally you may like to read the summary of Vieweg et al’s situational awareness paper that is attached to this article. It is a good example the first part of the Web Science cycle, where studies inform theory about a technology is used. The next part of the cycle would be to use this new understanding to create more powerful analyses and tools.

References

1. De Longueville, B., Smith, R. S., & Luraschi, G. (2009). “OMG, from here, I can see the flames!”: a use case of mining location based social networks to acquire spatio-temporal data on forest fires. Proceedings of the 2009 International Workshop on Location Based Social Networks, ACM

2. Lee, R., Wakamiya, S., & Sumiya, K. (2013). Urban area characterization based on crowd behavioral lifelogs over Twitter. Personal and Ubiquitous Computing, 17(4).

3. Mao, Y., Wei, W., & Wang, B. (2013). Twitter volume spikes: analysis and application in stock trading. Presented at the SNAKDD ‘13: Proceedings of the 7th Workshop on Social Network Mining and Analysis, ACM

4. Sakaki, T., Okazaki, M., & Matsuo, Y. (2010). Earthquake shakes Twitter users (pp. 851–860). Presented at the the 19th international conference on the World Wide Web , New York, New York, USA: ACM Press.

5. Signorini, A., Segre, A. M., & Polgreen, P. M. (2011). The Use of Twitter to Track Levels of Disease Activity and Public Concern in the U.S. during the Influenza A H1N1 Pandemic. PLOS ONE, 6(5)

6. Tsou, M., & Yang, J. (n.d.). Spatial Analysis of Social Media Content (Tweets) during the 2012 US Republican Presidential Primaries. Cartography and Geographic Information Science, 2013

7. Tumasjan, A., Sprenger, T. O., Sandner, P. G., & Welpe, I. M. (2010). Predicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment. Proceedings of the Fourth International AAAI Conference on Weblogs and Social Media

8. Vieweg, S., Hughes, A. L., Starbird, K., & Palen, L. (2010). Microblogging during two natural hazards events: what twitter may contribute to situational awareness. Presented at the CHI ‘10: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, ACM