Can Twitter Be Used to Predict Seoul’s Next Mayor?

campaign.socialmetrics.co.kr

Earlier in the year, we saw how Facebook played a big role in the Singapore elections. We also saw that online discussion of an election could be aggregated and visualized in an engaging and informative matter.

Right now in Korea, market research firm Daumsoft has put together a web dashboard that might give an indication of which candidate could win the October 26 vote for Seoul mayor.

Daumsoft’s dashboard, viewable at campaign.socialmetrics.co.kr, lets you see the discussion that surrounds each of the candidates (Won-soon Park and Gyeong-won Na), as well as which one is more ‘popular’ in Seoul’s 26 districts.

But can this service be of any help in predicting a winner? I was a little skeptical, thinking that while it might be an accurate reflection of young people’s interest, it might not accurately portray the interest of society as a whole. When I asked a Daumsoft representative about this point, he explained:

I’ll be the first to acknowledge that Korea’s Twittersphere skews younger than the general population and is therefore not completely representative… More mature platforms like discussion forums have many times more users from a wider demographic swath.

But he pointed the a more intriguing possibility, that being semantic analysis of the discussion, which is something the company is capable of.

We see semantic analysis of social media as one of many tools for understanding what is important to people. As an increasingly larger segment of the population begins to use social networking services like Twitter, sentiment expressed through these services will become increasingly representative of how the general population feels.

He added that Korea only has about four million users currently, and that ‘mature’ platforms like discussion forums would likely reflect on a wider demographic. Daumsoft’s dashboard was mentioned by Korea’s Digital Times and JoonAng Ilbo, so at least in some way, candidate interest on Twitter is seeping into more mainstream media. Whether or not it can shape or predict an election, of course, remains to be seen.

Data geeks can look forward to the 2012 US Presidential Elections for even more projects like this. My personal favorite from ‘08 was Everymomentnow.com, which looked at media reports leading up to the election, and visualized them on a timeline. That was way cool.

[Source: Newswire]