2011 Torino

1) Annotated data

Each row has: tweet, date, time, language, keyword, keyword score, polarity, subjectivity, profanity, retweets. The RT set contains English tweets that have been retweeted at least once, it is only 4MB and ideal for testing. The file "ows.py" is a Pattern script with a couple of examples of how to mine the data (for example, top negative tweets, top dirty tweets, tweets containing "arrest", ...) The keyword score can be used to interpret the "uniqueness" of the tweet. The polarity is based on positive/negative adjectives. However, since each Occupy Wall Street tweet could be seen as somewhat negative/revolutionary, you might experiment with only considering polarity > 0.5 as being positive.

2) Raw data