Emotion Rating from Short Blog Texts

http://delivery.acm.org/10.1145/1360000/1357229/p1121-gill.pdf?key1=1357229&key2=3377561721&coll=ACM&dl=ACM&CFID=86785888&CFTOKEN=56499296

Comments:
none yet

Summary:
The article presented a study done on how well people can perceive emotion with just plain text.  The researchers recruited subjects to analyze some text and mark which emotions were put into the text.  Some subjects were naive and some were considered experts at examining the text.  The goal was to determine what qualities in the text would help the naive reader analyze it closer to what the expert reader may see.

They found that the best indicator of how close the naive reader would be is the length of the text.  Shorter texts presented more challenges to the naive reader, presumably due to the smaller amount of clues about the writers emotions.  The researchers also noted that naive readers also were better suited to matching with the experts on joyful emotions.

Discussion:
This study seemed a little boring to read however I get the feeling that it was only boring because it was clearly a preliminary study.  The researchers are most likely looking at creating an automated system to analyze texts form for example the blogosphere, twitter, Facebook, or any other social application.  To this end, this study would be very useful.  An automated system would provide some of the best feedback to companies they could wish to get.

If I am right in guessing that the goal is to eventually create such a service I would be very supportive of their goal.  I feel that something like this, while difficult, is somewhat overdue.  It would be really nice to see this as a solved problem.