February 5th, 2008
We chose five different Super Tuesday states to gauge—California, Colorado, Georgia, Massachusetts, and Missouri. For these states, we used the methodology of combining activity and sentiment of state and national blogs.
We also made an overall prediction using a similar, but untested, method. We measured the sentiment and activity for national blogs. The candidate’s name had to be mentioned within a certain number of words of any one of the Super Tuesday states. The search parameters were necessarily looser than with individual states, but we’re hypothesizing that the “collective intellect” will outweigh any small inconsistencies.
(click to enlarge)
With the exception of California and Georgia, it appears to be a tight race. Massachusetts, Missouri, and Colorado are all separated by a mere 2 percentage points, which we would have to consider too close to call.
(click to enlarge)
For a full report on these predictions, detailed analysis and projections for the GOP races for the five states, and a three-month candidate sentiment report jointly researched with Marketing Pilgrim, please download this free pdf file.
To quote: “The candidate’s name had to be mentioned within a certain number of words of any one of the Super Tuesday states.”
This is a flawed way to determine sentiment, called “proximity searching” will only serve to generate inaccurate conclusions. Similiar to key word searching, also woefully inaccurate, none of these techniques can accurately determine sentiment-they do not understand “word meaning.”
Semantec NLP is what ’s needed to produce accuracy of intent and sentiment.