Alastair J. Gill, Darren Gergle, Robert M. French , Jon Oberlander (2008)
Emotion Rating from Short Blog Texts
Proceedings of the ACM Conference on Human factors in Computing Systems (CHI 2008)
Note: Arguments in this very relevant but then based analysis on blogs, which aren’t really a good example of online interaction.
P1
‘CMC, email, text chat and websites all offer reduced media richness’
‘different theories proposing varying degrees to which it is possible to understand social information’ e.g. diminished social presence, context. The SIP ( time element) put forward – ‘in a CMC environment with potentially unlimited time, intercutors would be expected to derive the same perceptions as is possible in f-to-f communication, either by placing greater emphasis on existing cue ( linguistic features) or by developing new strategies such as emoticons.
Refs 6 & 15 for the classification of emotions
Review of studies that led to the present study
Hancock et al, 2007 – Expressing emotion in text base communication
‘Stooge express either positive (happy) or negative (unhappy) emotions to their naïve conversational partner without explicitly describing their (projected) emotional state. Naïve subjects ( the text chat partners) could accurately perceive their interlocutor’s emotion, and were less likely to meet the author’s of negative messages relative to positive ones. Additionally a linguistic analysis of the transcriptsnfound that authors portraying positive emotion used a greater number of exclamation marks, and used more words overall, wheras authors’ texts portraying negative emotion used an increased number of affective word expressing negative feeling and negations’ Punctuation features matched the self-reported strategies used by the portrayers of emotion’
Building on Hancock et al
Expand classification of emotion from positive/ negative to 8 categories ( fear, surprise, joy, sadness, disgust , anger, anticipation, acceptance
Feldman Barrett, L and Russel J.A. (1998). Independence an bipolarity in the structure of affect. J. Personality and Social Psychology, 74, 967-984.
Plutchik, R. (1994) The psychology and biology of emotion. New York: Harper Collins.
Corpus from short blog texts rather than extended interactions
No actors
Method
65 students, frequent users of email, rare users of blogs. All were naïve raters of emotion.
Corpus – authors kept blogs for one month. From this corpus an expert assistant selected the first 200 words of each post if they contained some emotional content or were neutral. The 135 extracts were then rated by experts based on the 8 emotions and also a neutral condition. From these 20 were selected as expressing strong and clear emotional content, based on all 5 expert raters agreeing on the emotion identified.
Then shown to students ( naïve raters) using activation-evaluation wheel ( ie valence as well as emotion) judges asked to rate ‘ how they perceive the author’s emotions ‘ but ‘ not to spend too long thinking about their answer’
Results
Strong agreement between expert & naïve for joy, disgust, sadness, anger , anticipation
Conclusion
P3
Some emotion can be accurately expressed and perceived in short blog excerps thus contradicting social presence as less media rich such as asynchronous CMC.
(KRO but blogs are not about interaction and therefore co-presence)
method does not distort expression of emotions
