Tweetminster blog

Pages: 1 | 2 | 3 | 4 | 5

Tweetminster Daily - April 8 2010

The Tweetminster Daily is a summary of the day’s top political stories and analysis around UK Politics. The summary isn’t curated or editorially controlled, but entirely built using data around the most shared and clicked on links on Twitter.

Issue 41 - 08/04/2010 - 18:00

Today’s most shared stories on Twitter around UK Politics:


Posted at Thu, Apr 8th 2010, 18:19
permalink

Tweetminster Daily - April 7th 2010

The Tweetminster Daily is a summary of the day’s top political stories and analysis around UK Politics. The summary isn’t curated or editorially controlled, but entirely built using data around the most shared and clicked on links on Twitter.

Issue 40 - 07/04/2010 - 19:00

Today’s most shared stories on Twitter around UK Politics:


Posted at Wed, Apr 7th 2010, 19:11
permalink

Tweetminster Predicts points to a hung parliament

Last week we launched Tweetminster Predicts, an experiment in predictive modelling that will compare mentions on Twitter with election results. The aim of the experiment is to study the correlation between word-of-mouth (online buzz around candidates) and election results. In each constituency represented on Twitter we count the most mentioned candidate in each of the analysed constituencies - giving a first past the post prediction to the candidate with the most mentions.

As outlined in the study’s methodology, during the course of the campaign we will be releasing updated figures showing variations in the data and to take into accounts further constituencies mentioned on Twitter and new candidates that sign up to Twitter during the course of the campaign.

Today we’re releasing the first of these updates: 

Number of constituencies mentioned & represented on Twitter (and variations since March 26th) - 384 (+8)

Predicted top-line national party breakdown and variation:
CON 36% (+2) LAB 33% (-2) LDEM 22% (nc) Others 9% (nc). 
If these figures were repeated at the General Election they would lead to a hung parliament with Labour short of 26 seats.

Number of constituencies with more than one candidate represented on Twitter: 324 (nc)

Constituency-level predictions for these 324 seats: CON: 126 seats (+5), LAB: 135 (-3), LDEM 54 (-1), Others 9 (-1)

The main variations, although minor, have mostly been from Labour to the Conservatives. The Lib Dems have ‘lost’ a seat since our previous prediction, yet their topline figure remains unchanged and they’re trending positively. 

Constituencies which are displaying interesting trends this week include: Hampstead & Kilburn, Chippenham, Morley and Outwood, Glasgow Central, Mid-Dorset & North Poole, Manchester Withington, Stourbridge, Sutton & Cheam, Westmorland & Londsdale, Brentford & Isleworth, Luton South and Brighton Kemptown. 

We have also updated the Tweetminster Predicts paper outlining the study. Our website will be updated on Tuesday morning.


Posted at Mon, Apr 5th 2010, 17:45
permalink

Visualising sentiment trends more clearly

Alongside, launching our Election platform, today we’ve also changed the way we display sentiment scores - shifting from a -100 to +100 range to a 1-5 scale where 5 is highest positive sentiment.

We believe that this change will make it much easier to display how trends vary over time as it provides a much cleaner (less spiky) graph when the data is plotted. For example, comparing sentiment around the three main parties over the past month displays a far clearer narrative.

In addition to the month-over-month view in Search, our Election platform displays near-real time sentiment scores that update every minute via a ticker.

Finally, a tip - analysing sentiment is different to polling-information - direct comparisons between parties for example aren’t necessarily always interesting, far more insightful trends are in the variations within each party and comparisons between variations.

Sentiment is one metric that adds a layer of value to a holistic set of insights & data that we’re powering at Tweetminster - it’s in fact always helpful to view sentiment in the context of other indicators, including volume, associated terms and relevant content. 

Here, you can read more about how Tweetminster’s sentiment analysis works.


Posted at Thu, Apr 1st 2010, 17:54
permalink

Pages: 1 | 2 | 3 | 4 | 5