This study and this presentation are fascinating examples of what’s occurring in sentiment and message analysis. The first study focuses on a sampling of 9,664,952 tweets from 2008 and found that Twitter (surprise ;-D) is an excellent media for analyzing public sentiment on topics of national significance (e.g., the 2008 Presidential elections). The authors find that social, political, cultural, and economical issues have relatively short-lived effects on public mood, but that Twitter may “highly affect the dynamics of public sentiment.”
The presentation cited above gives an overview of the concept of media influence modeling to better predict audience opinion formation. The author cites as theoretical underpinnings: Agenda-setting theory, opinion leadership theory, social influence theory, co-orientation behavior, priming and framing theory, and information theory. What’s interesting about this presentation is the depth and sophistication going into analyzing message influence. The bibliography (page 13) is an excellent resource too.