wc_20181220 1450.png

The Brexonometer

A live barometer of how the nation is feeling about Brexit 

Rolling Sentiment of ‘Brexit’ Tweets


Welcome to the Brexonometer

The Brexonometer was created on 7th December 2018 to give moment by moment insight into how the nation is feeing about Brexit by applying advanced analytics to compute the real-time sentiment of Tweets.

At the time many folks felt that the country was getting Brexit fatigue. Judging by the sheer volume of Tweets on the subject this is clearly not the case. Since its launch as an inconspicuous Blog article The Brexonometer has grown in popularity to the point that we needed to create this dedicated site.

Live Brexonometer sentiment scores are published here every day. We hope you enjoy following its ups and downs as much as we do and pop back occasionally as we’ll be adding new features over the next few weeks.

How to Read the Diagram

The Brexonometer is a barometer. Barometers don’t tell you what the weather is going to do. In fact they don’t even tell you what it’s doing right now. But they do provide useful information. Is the pressure rising over 1030? That probably indicates an Atlantic high is bringing warm weather and sunshine. Did the pressure suddenly drop below 900: then there is a BIG storm just around the corner. The Brexonometer is the same, providing a view of the overall sentiment across Twitter. It won’t tell you how a second referendum would go. It won’t tell you how the next Parliamentary vote will go, but it will tell you loud and clear that people got really mad when a bunch of yellow jackets held up traffic in London.

The Brexonometer is impartial. It simply measures the aggregate sentiment within the language used in each batch of Tweets. The solid lines show the trends: these tend to stick around zero until something big hits the news wires. Then the public response is often very apparent as you see sudden changes in sentiment, either positive or negative. Remember the system does not take sides, but does shine an unbiased light on the public’s mood in response to emerging events.

How Does it Work?

All English language Tweets mentioning the word Brexit are aggregated into mini-batches of messages. We then use natural language processing technologies to compute the overall sentiment of each batch. This is tracked continuously over time. The latest live values are shown in the graph above.

Values over zero indicate more positive than average sentiment while values below zero show more negative than average sentiment. Samples are plotted sequentially to create a time series of Tweet sentiment (purple dots). Each dot represents a mini-batch of Tweets so at busier times dots become more frequent.

The time series on its own is very noisy and trends are difficult to discern so we filter the raw feed using moving averages. Averages taken over 10, 30 and 100 sample sliding windows are shown with solid lines. These provide fascinating insight into the ‘mood’ amongst Twitter users as it changes en-masse over time. 


What’s Everyone Saying?


Words most frequently occurring in ‘Brexit’ Tweets over the previous hour.


More Feedback Please!

The chart above just counts the number of times words have appeared on Twitter in conjunction with the word Brexit over the proceeding 60 minutes. The top 200 most frequently occurring words are then crammed into the diagram, with the font size reflecting the word occurrence count. It really is simple, but does afford some insight into what everyone is Tweeting about. We’ll update this trial chart every 10 minutes during the day.

One obvious improvement would be to give folks the option to rewind to a certain point in time so they can see how the spread of words changes through the day. We will consider this if enough people ask for it. Tweet @breonometer if you would like to see this feature added.

People have asked for the Brexonometer to be extended to show the spread of opinion associated with Tweets. The Brexonometer works on large numbers of messages gathering an overall consensus. The algorithm and methodology are straightforward (it’s free remember) but work over large numbers of Tweets as oddities average out. Opinion Mining is a bit tricky for individual Tweets and / or smaller populations. We have a more advanced version. If you are interested in finding out more please contact us.

Thank you for your interest so far and please follow @brexonometer and share the link if you like what you see.

The @brexonometer team.