When I did some work on sentiment analysis using tools like Crimson Hexagon, it always felt like the data was skewed since what people post is different from what they say, which itself is different from what they do. Might need to be some kind of corrective filter that includes 'uncounted emotion' as a baseline.
I mean the elephant in the room is that that "sentiment analysis" is flawed and crude and does not correlate in any meaningful way with actually "happiness".
http://wefeelfine.org/methodology.html
When I did some work on sentiment analysis using tools like Crimson Hexagon, it always felt like the data was skewed since what people post is different from what they say, which itself is different from what they do. Might need to be some kind of corrective filter that includes 'uncounted emotion' as a baseline.