How does it work?

Tweetminster Search measures the pulse of UK politics in real time - it shows you how people feel about the issues of the day - as expressed through the thousands of messages posted to Twitter by politicians, news source, journalists and the public - and cuts through the noise to help you find relevant content. Because our search works in real time it shows what the people think now and tracks how their opinions and news stories have changed over time.

It's easy to use. Enter a search term (eg. Iraq Inquiry) or up to three multiple terms (eg. Labour, Tories, Lib Dems) and click 'Search' - the results show you:

  • The number of relevant posts about the subject over time: our system dynamically creates a sample from thousands of people that is weighted, and is quantitatively and qualitatively representative of Twitter as a whole.
  • How these posts break down into positive, negative and neutral opinions (sentiment) - the sentiment score is a quantitative measurement, on a scale of 1 to 5, based on average sentiment determined by our algorithms within content from the analysed sample.
  • Plus lots of related information - associated terms and issues, links to most shared URLs, the individuals with the most impact on the subject (influencers) and a location breakdown for where the terms have been discussed the most.
  • And relevant content and trends.

The results give you an at-a-glance breakdown of how big a topic is, what people think about it, how many people are interested, who is driving the public debate, how it all fits together with the other issues of the day and the sources that back it up.

Alongside terms, you can also analyse Twitter usernames (eg tweetminster - i.e. without the @), or a combination of elements. For example you can analyse the correlation between a topic (e.g. Tax cuts) and user (EricPickles) or search for a term within a party or amongst the tweets of MPs only, by using the advanced filters listed under the search bar.

Our systems are designed to only analyse and present relevant content which is what makes Tweetminster Search different. We define relevance by determining the influence, authority and reach of a post and coupling this with dynamic sampling that is representative of Twitter as a whole.
Our analysed content is fed from the people our network analysis has identified as making up the UK politics network - this means politicians, journalists, bloggers, news sources and the people they're having conversations with. These influencer tweets create the core sample for all the data we serve and as new people become relevant through interacting within this network we can identify them algorithmically meaning our politics network is always up to date.

We'd love to hear from you, send your feedback and ideas over Twitter: @tweetminster.

How it works

  1. The premise of our methodology is a network analysis of politicians, media, influencers and "members of the public" that are on Twitter and post about politics - this network is in the high thousands, and is constantly growing.
  2. When someone does a search on Tweetminster we create a sample from our network that is dynamically weighted and representative of overall volume and population of tweets for that search over a period of time (i.e. each day the size of the sample will also vary depending on how popular that term has been on different days).
  3. This means that in terms of tweets from MPs, PPCs, journalists, news sources and members of the public, the sample will be relfective of volume, affiliation (when applicable) and conversations/tweets across the network.
  4. In doing so, we look at tweets that have greater reach, influence (their position is "echoed" by many), and only count orignal tweets and not the retweets it generated.
  5. Our analysis is semantic: Brown is not a colour and Cameron isn't a film director. So while a Twitter search for Brown or Cameron may give 1000s of results, we may only analyse a sample of 100s of tweets - these will though be reflective of specific meaning (Brown and Cameron the politicians) and representative of the total volume and opinion of tweets within the network.
  6. Our algorithms then determine sentiment by looking at positive, neutral and negative linguistic meaning within tweets. The sentiment score is a quantitative measurement, on a scale of 1 to 5, based on average sentiment within content from the analysed sample. A score of O occurs when positive and negative balance each other out, or all content is neutral.
  7. It's worth noting that volume and reach are equally important - for example if an event has greater reach and resonance amongst one party over another this will be reflected in the sample, and consequently in the results.
  8. All the above happens in near real-time.
Tweetminster.co.uk | © 2009-2010 Tweetminster | Follow us on Twitter: @tweetminster