BBC BLOGS - dot.Rory
« Previous | Main | Next »

The chancellors, Twitter and sentiment analysis

Rory Cellan-Jones | 09:42 UK time, Tuesday, 30 March 2010

I don't quite know what came over me yesterday evening - but as I headed home I found myself looking forward to a live television debate between three politicians.

It wasn't so much the cut and thrust of the argument between Alistair Darling, George Osborne and Vince Cable which quickened my interest, but the prospect of the first big social-media event of the election campaign.

Alistair Darling, Vince Cable and George Osborne

For the last year or so, major television events - from big football matches to reality shows like The Apprentice - have become a lot more interactive because of the social-networking activity that happens around them, mainly on Twitter.

That's started happening with political programmes - from the appearance of the BNP leader Nick Griffin on Question Time to last week's Budget.

But the three prime ministerial debates during the election campaign will bring all this activity up on to a whole new level - and the chancellors' debate was a dress rehearsal for that.

I watched the programme while scanning Twitter and Facebook, and it did add an interesting dimension to the big fight.

Every feint, every parry, every joke (not too many of them), was analysed, applauded or booed by the online crowd. What was immediately apparent was that Twitter, not Facebook, is the arena for this activity.

Facebook is trying to whip up interest in politics through its Democracy UK page, but its page on the debate only attracted a few dozen comments, while there were more than 12,000 tweets during the hour-long debate.

Each candidate had his own claque of virtual cheerleaders determined to call the debate in his favour.

So after a while, the Conservative chairman Eric Pickles tweeted: "George Osborne clearly won the first half hour of the debate."

Vince Cable's fellow Liberal Democrat Susan Kramer chimed in with: "Vince brilliant but amazed that George is most dull!"

And the Foreign Secretary David Miliband added this view from across the Atlantic: "Just arrived canada for 24 hours for g8 foreign mins meeting. Heard Alistair won the economy debate."

But there appeared to be plenty of real people as well as party spinners joining in, and it was the Liberal Democrat candidate who edged it in terms of online approval.

Channel 4 ran its own internet poll during the debate, with Vince Cable getting 36% of the votes, and Alistair Darling and George Osborne tying on 32%.

The guidelines for this encounter were loose enough to allow a measure of real debate between the three men, skilfully moderated by the presenter Krishnan Guru-Murthy.

The rules for the prime ministerial debates are so many and so restrictive that there may be a lot less interaction between Messrs Brown, Cameron and Clegg - and that could make the social-media activity around the debates even more compelling.

Twitter, rather than television, could be the place where the issues are really dissected.

But is any of this important, do the tweets or blog comments or Facebook updates of thousands of voters provide any useful guide to what the electorate as a whole really thinks?

Well, some are betting that it does matter. They believe that careful analysis of this activity can deliver data which could be as useful or accurate as opinion polls.

In the last few weeks, I've been hearing a lot about something called "sentiment analysis", a branch of computer science that attempts to extract meaning from words expressed online - you can read a useful Wikipedia definition here.

The big challenge facing this science is that computers just don't do irony - and when it comes to analysing the sentiments of British voters, that's a problem.

But a number of companies are planning to use the technique to try to look into the minds of the online crowd during the election.

An organisation called Tweetminster, which analyses political activity
on Twitter, is using another term, "predictive modelling" - for an interesting experiment it's planning to undertake.

It will track the online buzz around candidates and constituencies in the forthcoming election and work out whether there is any correlation with the winners.

Alberto Nardelli from Tweetminster says a similar exercise carried out during Japan's general election last year proved accurate in predicting the results of 90% of constituencies.

Well, we shall see. After all, sentiment analysis or predictive modelling of last night's debate might lead you to conclude that the next chancellor will be Vince Cable, but I think you could still get quite good odds on that from the "sentiment analysts" at Ladbrokes or William Hill.


  • Comment number 1.

    Interestingly, yesterday's data points to a different winner other than Vince Cable

  • Comment number 2.

    Any serious company director or manager, had they been confronted by this trio as candidates for a middle-ranking Finance Director post, wouldn't have hired any of them. Caledonian Comment

  • Comment number 3.

    @ 1. Well only if you count how many times someones name was mentioned. Im sure they weren't all positive mentions!

  • Comment number 4.

    One of the big problems with viewing Twitter as a way to analyse the results is that it's about the size of the Twitter 'team' voting for each contender. Most of the people that will decide the election weren't on Twitter and possibly weren't even watching.

    I did both - and thought that Osborne won -

  • Comment number 5.

    It's an interesting (if not yet scientific) field — for example I've been monitoring social web content produced in Birmingham for almost two years now and producing an "emotional wellbeing" rating (as it's based on any sentiment rather than around a subject). It does rise and dip around weather and sporting events as you'd predict — have a quick look on twitter (@birminghamuk).

  • Comment number 6.

    "Twitter, rather than television, could be the place where the issues are really dissected."

    It could be, but it won't be.

    Twitter's brevity is its strength and its weakness. Issues won't be dissected, but style and impact will be, mercilessly. How many comments were there on their ties vs their economic policies?

    There will be a lot of chatter on Twitter and that is great - but it will primarily reflect the opinions of those tweeting rather than impartial dissection, as has always been the case.

    Unfortunately, television debates are all about style and how you come across, and very little to do with substance. Remember the West Wing episodes about debate prep?

  • Comment number 7.

    Technicalfault: "How many comments were there on their ties vs their economic policies?"

    I see you went to the trouble of making an analysis. Thank you for your effort.

  • Comment number 8.

    "At 11:06am on 30 Mar 2010, Caledonian Comment wrote:
    Any serious company director or manager, had they been confronted by this trio as candidates for a middle-ranking Finance Director post, wouldn't have hired any of them. Caledonian Comment"

    Vince Cable was chief financial officer at Shell in the mid nineties..

  • Comment number 9.

    @1. I don't quite see how being mentioned the most is a good thing, as most of the tweets I've seen mentioning "George Osbourne" OR "George Osborn" have been massively critical of him.

  • Comment number 10.

    Watchable but pointless, given that none of the three will ever get to be Chancellor.

    Real debate should be between Balls sitting on Brown's knee and Clarke.

  • Comment number 11.

    Sorry to blow a big hole in your digital election thing Rory, but Twitter users are not a cross section of the population. Those who tweet on debates such as last night's are invariably politically active and generally have strong party allegiances already. A great many of the tweets made during last night's event would have been from party activists. As such, the tweet levels are related more to the organisation and tech awareness of each party machine than anything else. The tweeters certainly aren't the floating voters in marginals who could actually make a difference to the election result.

    The tweeters don't influence other twitter users. The only people they seem to influence are journalists. What journalists then report about twitter activity is more influential than the tweets themselves. But as I said, the tweets bear little relationship to general public sentiment. An opinion poll, despite its flaws, is always going to be far more representative than an unscientific analysis of an unrepresentative sample.

    Despite all the technology tools available, this election will still be won and lost on the TV and the print media.

  • Comment number 12.

    Rory - I followed the debate over the internet and tested what happened after voting in the Channel4 poll.

    Do you know if the IT bods at Channel4 are just using cookies to control/prevent repeat voting?

    If so then I'd suggest their internet online poll has zero credibility because of the risk of multiple votes by those who know a smidge about how their PC works.

    I'm happy for this post to be zapped after you have read it - don't want any libel actions do we! :-)

  • Comment number 13.

    Having heard three unconvincing prospective chancellors, I declare that the victor of the debate is ... the IMF!

    After all, given the unconvincing arguments we heard, I expect the IMF will be running the economy shortly after the election.

  • Comment number 14.

    This comment has been referred for further consideration. Explain.

  • Comment number 15.

    I think using sentiment and measuring social media buzz around a particular topic is a great way to take a proverbial pulse of how the public is feeling (for example, at Biz360 we are measuring social media conversations around American Idol vote-offs

    However... to echo the sentiment of some of the commenters on this thread: not everyone is on Twitter, and although a lot of people are on it (especially in the U.S.), it skews a lot more tech-savvy and higher educated the the general population. When using social media as a proxy for how the entire country feels, do keep this in mind.


    Maria (@themaria on Twitter)

  • Comment number 16.

    For me it feels like we are on a verge of a new revolution. Just as the moving picture was added to audio broadcast we are seeing the emergence of a new channel of communication to augment television broadcast. You can see in the research the BBC is doing on subtitles over Internet that it is only a minor leap to integrate services like twitter. And the fact that Google are exploring a similar avenue with their TV box can only but add weight to this development.

    An earlier comment by Technicalfault mentioned that "twitter's brevity is its strength and its weakness. Issues won't be dissected", this doesn't have to be the case as the back channel discussion and the broadcast can easily be re-aggregated allowing people to review and analyse the two streams in context (as demonstrated here

  • Comment number 17.

    Undoubtedly one of the best places to watch for the view of the crowd is Betfair, where people bet against each other to set the odds. Thus, you get real crowd opinion and not just some view expressed by a bookie.

    The markets there react almost instantly to any event and in the wake of the Channel 4 debate last week it was George Osborne and the Conservatives who punters were backing on the site, and you could see that all live in real time.

    They have also launched a site in partnership with ComRes, a polling company, that shows live betting and the comparison with polls. It is at and well worth a watch.

    Twitter is good for this kind of thing but to get genuine sentiment of what people think (versus what people want) it's always better to look at the thousands of people who are parting with their hard-earned cash on Betfair. 15,000 people bet on the US election on Betfair and it'll be a lot more this year for the General Election here.

  • Comment number 18.

    This comment has been referred for further consideration. Explain.


BBC © 2014 The BBC is not responsible for the content of external sites. Read more.

This page is best viewed in an up-to-date web browser with style sheets (CSS) enabled. While you will be able to view the content of this page in your current browser, you will not be able to get the full visual experience. Please consider upgrading your browser software or enabling style sheets (CSS) if you are able to do so.