Conservative Party are the best tweeters as electioneers turn to Twitter for votes

May 5, 2010 | by | 0 Comments

The Conservative Party were praised as the ”best tweeters” as it emerged that a FIFTH of politicians ”tweet” out of line with party policy.

Computer program TweetGuv, created by science students at Cambridge University, automatically analyses and compares politicians ”tweets” on social networking site Twitter.

Twitter’s 140-character limit format has been adopted by many politicians who see it as a way to boost their popularity and make themselves accessible to the electorate.

TweetGuv analysed the tweets of 126 conservative MPs and PPCs, 159 from the Labour party and 112 Liberal Democrats.

It plots tweets into a ”triangle of alleigance” graph to show the extent to which each MP is towing the official party line on the social networking site.

Of 75,449 tweets analysed, 39,889 were deemed to be political in nature and of these 28,854 – or 72 per cent – were in line with party policy.

Conservative tweeters fared the best with 84 per cent of tweets towing the party line according to data collated from the project.

Tweets from the Labour Party were 79 per cent in line with policy while Lib Dem tweets were 78 per cent in line with policy.

The programme also profiles each individual politician classifying their tweets as being in line or slightly in line with the policies of one of the parties, not necessarily their own.

It was fed with 205,152 words of training data from party political speeches and broadcasts and parliamentary debates.

TweetGuv was conceived as a group project undertaken by six undergraduates in computer science at the University of Cambridge.

Oliver Chick, 20, of Gonville and Caius College, said that it had provided ”some interesting results”.

He said: ”The idea of mapping politicians’ tweets algorithmically to their party’s official line was actually given to us by a lecturer in the Computer Laboratory.

”However the rest of the design was decided upon by us.

”Experimental results show our party classifications are correct 68 per cent of the time, for political tweets.

”We have built a series of mathematical models to be able to detect noise in wrong classifications and deal with this accordingly.”

Other collaborators were Sebastian Probst Eide, 25, of St Edmund’s College, Harry Askham, 19, from Christ’s College, David Reid, 20 of Churchill College and Jason Bell, 20, of Robinson College.

Another collaborator, Frances Zhang , 21, at Trinity College, is taking part in the MIT Exchange Program Massachusetts Institute of Technology.

Spin doctors from all parties have realised potential to connect with the voter, but also to form opinion as events happen.

During its broacast Channel 4’s debate Ask the Chancellors was the fourth-biggest ”trending” or talked-about issue on Twitter anywhere in the world.

At its height, there were about 200 tweets a minute. Over the whole hour from 8pm until 9pm there were 15,000 tweets and about 5,000 in the hour that followed.

Two notable exceptions to the Twittersphere are David Cameron and Gordon Brown although Sarah Brown is a keen Twitterer with more than 1.1 million followers.

Lib Dem Leader Nick Clegg is a keen Twitterer with over 5,000 followers as is Vince Cable who has 4,500 followers.

But politicians have also become aware of the dangerous potential for their accounts to hacked or their voices impersonated.

Red-faced Energy Secretary Ed Miliband posted a link to a website with the message ”Hhey [sic], i’ve been having better sex and longer with this here” on 26 February.

But hours later he made it clear he had been hacked, posting: ”Oh dear it seems like I’ve fallen victim to Twitter’s latest ‘phishing’ scam.”

The six-week project has now finished but TweetGuv’s creators will add more training data to update and improve the programme’s accuracy.

View TweetGuv online at

Tags: , ,

Category: Uncategorized

Add your comment

Libellous and abusive comments are not allowed. Please read our House Rules

For information about privacy and cookies please read our Privacy Policy