Can Twitter sentiment analysis predict outcomes (like the Irish election)?

When I was growing up, election coverage was characterized by an exuberant political pundit leaping around large cardboard charts of the UK with the kind of coloring normally reserved for the weather report. The ‘exit polls’ we were familiar with only updated about every four hours and only included those people who were prepared to be cornered by the political researchers hanging around near polling stations.

Fast forward to 2011.

We currently have a general election unfolding in Ireland. The Irish online news site The Journal has been crawling over Twitter, that political social network du jour, using the conversations that happen there to predict which way the election will sway. And so far the headline graphic looks like this:

image

It’s a great case study in the current status of analytics and throws up some wonderful points that have relevance beyond the Irish political scene.

Data is everywhere

Researchers no longer need to go in search of data. Whilst I don’t deny the added color and in-depth insight from questionnaires, focus groups and other tools used by human researchers (whether in the political or commercial realm), there is rich data out there that you don’t have to force out of people. Social networks like Twitter and Facebook give us access to voluntarily-provided information on social groups. We no longer have to bug people to provide us with data.

Growing importance of social media analysis

Let’s face it, we’ve seen a huge growth in the use of social networks over the last two years (not sure why I pick that time frame, maybe tied up with when Twitter/Facebook buttons first starting appearing in ads and on TV). We’ve taken our social lives online. And the beauty of being online is that everything can be tracked. We leave traces. and when you aggregate all of these, patterns start to appear. Is this level of analysis creepy? The privacy issue definitely has to be considered, however I’d contend that the information is so much more valuable in aggregate (effectively anonymized) than it is at the individual level.

Sentiment analysis can throw a curve-ball

Here is what the volume of conversations around the Irish election shows us:

image

Now look at the sentiment:

image

Fine Gael have by far the most conversations. However, much of this conversation is not positive. I’d say from a marketing perspective this is something we need to pay more attention to. Far too often we’re still using raw numbers as a determinant of campaign success. We need to add the sentiment layer on top to understand more of the nature of the conversations we ignite.

Presentation is everything

The first image I highlight in this post is so immediately descriptive. Newspapers have been producing wonderful infographics for decades. In the business world we still end up with reports that look more like this:

(not meaning to pick on anyone, this is just an image that came up in a search)

How much further our story goes if we take time to package it up. Business analytics will only move further into the mainstream if the findings are presented in an easily-consumable fashion.

So, having stuck my neck out in favor of The Journal’s Twitter Tracker, I’ll have to come back next week with some post-election analysis. In the meantime, back to Twitter to watch this election unfold.