In a paper I published last year, I introduced a novel algorithm for sentiment detection in movie reviews. I used data from Metacritic, which consists mostly of brief reviews (only a few sentences). In that sense, they are pretty similar to tweets. So I thought, why not try the algorithm on data from Twitter's streaming API? The idea was to follow a developing story and graph the sentiment expressed by the tweets about it. I decided that a football game is a perfect scenario for that, as the sentiment can change from positive to negative over time (one or the other team leading). To use the streaming API, you have to chose a keyword, and it will give you all the tweets about it. On Dec. 13, 2009 the Eagles played the Giants, and I chose "Giants" as the keyword. I'm not a fan of either team, Giants just happened to be more trending than Eagles. So here are the results:
The graph is a running average of the sentiment of the tweets over time, with a window size of 999. The sentiment is generally positive (> 0), probably explained by a lot of cheers ("Go Giants!" etc.). It drops towards the end of the curve, which makes sense because the Giants lost that game.
This is just a very brief analysis and there is a lot more you can extract from the graph, e.g. match the spikes with scorings. I'll update as soon as I have more results.