Monitor Setup

On SWI server, create topic list "Governor Election 2010" and then add a topic for each governor candidate for all the state races.  The candiate names are from Politics Daily.

Monitoring

Twitter conversations and blogs are collected from Twitter and the blogsphere.

Categorizing Social Conversations

Set up categories relevant to election:

• Positive sentiment
• Negative sentiment
• Issues
• Campaign
• Unknown
• Exclusion
  

For each candidate, tweets are clustered for each day starting from 10/27/2010. Read through the clusters and manually assign a category to each cluster read. For those conversations with clear feature such as key statement or phrase, copy and add the feature to the category assigned.  As you manually assign categories, SWI server is learning from the features in real time and does automatic categorization after each new feature is added.  This automatic categorization helps you categorize the tweets faster.

Finding

California race: The sentiment difference is startling on Twitter. From the randomly selected tweets since 10/27, the ratio of positive to negative sentiments is 1.71 for Jerry Brown and 0.32 for Meg Whitman.