Saturday, 19 February 2011

Web 3.0 Lab: Twitter analysis predicted demos in Pearl Roundabo...

Overview

By analysis the tensity and pattern of geo-location tagged tweets in Bahrain we were able to confidently predict a mass demonstration in Northern Bahrain several hours before it began.

Our Prediction

Based upon our experience and analysis of the twitter trends on the ground in Bahrain this morning, we were able to make this prediction at around noon Bahrain time:

Web 3.0 Lab: Anticipating more protests in North Bahrain today:

"From our morning analysis of the intensity of tweeting in North Bahrain, the pattern and spread of this intensity, and the content of #Bahrain and #lulu hashtag on Twitter we are predicting that there is a high probability of continued civil unrest and protest in North Bahrain, and that the Regime will again likely not be able to contain or control the scale or extent of protests."

Track tweeting from around Lulu Pearl Roundabout in real time with this tool.

This is significant because at that time news reports we were reading from journalist inside Bahrain were saying that it was uncertain what would happen. We were reading some reports of people being scared after recent killings and even of smaller turn out.

This contradicted what we were seeing on the ground with our twitter analysis tools. We could see a massive crowd of people who were twitter so much that there must have been 10,000s of them. They were mass in high concentrations near Bilad Al Qadeem. At one point this morning tweeting levels from that location were almost as high as those in the center of NYC.

How we made the prediction

On previous days we had seen this pattern before mass protests on Pearl Roundabout. We could tell that a massive number of people were collected and were motivated to communication. Our observations of Bahrain people living in London is that protests involve heavy levels of social networking. Protesters in Bahrain are making extensive use of various social networks to coordinate and inspire protesters, and to keep in touch with large families.

People in Bahrain have far more extended families than most westerners. People preparing to undertake a risky protest against a violent regime would also want to keep their family informed. Twitter is an ideal tool to broadcast status to a large group of people.

So from these factors we ignored the cautious statements from major media sources and predicted that a mass demonstration would take place in Northern Bahrain and that the regime would not be able to control it without resorting to mass killings.

The protesters would have had a mixture of deep determination and concern for their safety and the safety of their families. We have spoken with many people from Bahrain who struggle constantly to keep in touch with their extended families. Social Networks on the Internet work well for this.

So this spike west of Pearl Roundabout could only mean a group preparing to confront the regime.

Several hours later the news reports came in that this is precisely what happened.

Limitations of our current model (and yes we are working on these)

But there were certain things we could not see with our tool. We could not see the retreat of the Army for the Roundabout or peaceful strategy protesters would take. We continue to explore semantic analysis to give us a more nuanced understanding of what large mean.

But we are pretty confident that we have demonstrated the value of this kind of analysis, and this will be part of the news gathering and reporting of the future.

Also our model has been useless in Libya, were events are moving very quickly. Our analysis of Libya has been based on semantic analysis of the #Libya. This has lead us for some time to believe the regime there was weaker than many people imagined in the west, but we are still developing the solid model to make more firm predictions based on Twitter content in this area.

Conclusion

By using our Twitter density analysis model we were able to see people massing this morning west of Pearl Roundabout. From our previous observations and our analysis of the related Twitter and Facebook content we were able to confidently predict a highly probability of a protest starting today in Northern Bahrain. Several hours later this is what happened.

We feel this has helped demonstrate the value to journalists to this kind of analysis in the right conditions. There is a requirement that a society have a high level of Internet adoption. Our tool has been useless in trying to see what could be happening in Yemen and Libya. But for Bahrain it has proven very effective.


Web 3.0 Lab: Twitter analysis predicted demos in Pearl Roundabo...:
Post a Comment

Official Linden Blog