Emotions play a vital part in our daily research here at Liechtenecker. We constantly try to understand people, feel their sentiments and thus create empathic interfaces tailor made for them. Football often raises the best as well as the worst emotions in us to the surface. A perfect playground for our research!
We asked ourselves: How and what does Europe feel during the soccer championship 2016? Is there any way to get significant data between and during the games?
The idea
Twitter is the ideal testbed for our idea which is a realtime map of all the different emotions in Europe during the one month tournament. It provides a very concise, unfiltered outlet for its users sentiments, thus being the best stage for our project.
Getting the data
At first, we obtain the data from twitter. What sounds like an easy job requires a lot of preparation: there is no official hashtag for the tournament (although #euro2016 is widely used), so we monitored tweets well ahead of the tournament to gather a list of all football-relevant hashtags. In addition, we added all game pairings (e.g. #AUTHUN) as well.
Processing
First in the pipeline is our tailor-made emotion recognition. We tried a few different approaches, but settled on a combination of evaluating phrases and emojis in every tweet. Tweets that have no strong emotional context (or none at all) are discarded – these are approximately 90% of all incoming tweets.
Next up is geocoding. Most tweets are sent without geolocation enabled, so we have to manually extract this information from a users profile. All tweets that are outside of Europe are scrapped in this step.
PubNub Backend Interface
The remaining tweets – neatly tagged with their emotion as well as geocoded – are then sent to PubNub. PubNub offers an incredibly fast realtime messaging system and connects our backend to our frontend. Using PubNub was an easy choice, since we already successfully used it for our HypeCompass. It offers an easy to use API for backend (php) and frontend (javascript), thus making the implementation easy as cake.
Displaying the data
We chose a relatively abstract version of displaying the data by representing the european land masses by dots. Every incoming tweet is assigned to its closest dot on the map, the color changes respectively to its emotion and all neighbouring dots are – to some extend – influenced as well.
The finely tuned map visualises the different emotions of europeans during and between the games.
Learnings & the future
The tournament isn’t even over yet and we’ve already learned a lot. Handling this much data puts a lot of stress on our server (both cpu and storage). During the first games, we came to realize that we underestimated the number of simultaneous tweets. All tweets are processed sequentially one after another and that could lead to a delay of up to ~30 minutes from receiving the tweets to displaying them on the map. Since then we tinkered with our implementation and server setup to drastically improve the app’s performance.
We’re still very much committed to the idea and are already planning for the next big international events. #esc? #rio2016? Stay tuned!
