Tommy Collison

Last Saturday I went along the BT Young Scientist & Technology Exhibition and was once again impressed by the scope and individuality of all the projects at the exhibition. One that particularly caught my eye was this one by James Eggers, measuring the mood of Twitter:

My name is James Eggers, Iʼm an Irish student living in Dublin, Ireland. Back in September of 2010 I started thinking about the BT Young Scientist and Technology Exhibition 2011. The Young Scientist Exhibition has been something Iʼve wanted to do for years, but unfortunately never got round to it. I made it my mission to do it this time round. Iʼm a big Twitter user (Iʼm @JEggers2), and so I decided that this hypothetical project would be based primarily around Twitter. Eventually, after much brain storming I came up with a cool idea for my project.

I decided to use Twitter to figure out how the collective mood of people in Ireland changes and fluctuates over time, but with an added twist. I wanted this experiment to be different. I decided to plot this information on a map of Ireland on a county-by-county basis and make a short time-lapse video, showing the mood of each county fluctuating on every minute of the day. I also decided to create a real time, auto-updating, live mood-map of Ireland.
With these objectives in mind, I started work on the project. First off, I put together a simple data miner, capable of mining about eighty thousand Tweets from Twitterʼs API per day. Then, I set about creating the necessary algorithms for the project. I needed something that would sort the Tweets into their respective counties, and to score each Tweet based on itʼs mood. Doing this properly took me the best part of two months.

Scoring the Tweets was problematic. Computers are terrible at understanding human language. I had to do a lot of research. I learnʼt huge amounts about Natural Language Processing and Semantics. I applied what I learnʼt to my algorithms, which worked very well in the end. A lot of these types of algorithms take shortcuts and donʼt work properly. For Example, some algorithms would take a sentence like “I am not happy” as being a happy sentence because the word “not” isnʼt taken into account. The algorithm failed to take the context of each word in the sentence into account. I built my algorithm to fix this problem, along with a slue of others, creating a very accurate algorithm.
In December 2010, I was pretty much finished with the programming. I had about four million Irish Tweets. I started putting my algorithms to work on these Tweets. When they were finished, I had some interesting results.
This is a time-lapse video of how the collective mood of each county in Ireland changes in Ireland over an average day:

I also created that real-time mood map, it works very well. You can take a look here: http:// realtime.thevibesofireland.com/. The more green a county is, the happier is. The more red a county is, the less happy. You can also hover over each county to get a live graph of the fluctuations in the mood of the county and some interesting statistics about each county.

The results of this experiment were very interesting. I found that on average people on the West Coast of Ireland are almost always happier than people on the East Coast of Ireland. I also found that on an average day people are happiest at about 18:00 and least happy at 04:00 to 08:00. Another interesting find was that people in Ireland were generally not happy about the budget (December 2010), but very happy that Ireland agreed to accept financial support from the EU.

You can also take a look at all my other findings and work at http://thevibesofireland.com/. Thanks for reading, hope you found this interesting!