SentiMetrix

Most of today’s sentiment analysis technologies focus on the polarity of opinion (good/bad, positive/negative), not on its intensity.  We believe this is insufficient.  Why?  

We track sentiment data in order to understand, and foresee people’s behavior.  This is not new: pollsters and marketers have been doing it for many years.  Except for simple, binary outcomes, such as elections exit polls, polling and marketing research typically involve fairly granular questions.  These questions gauge the intensity of the respondents’ sentiments by offering multiple choices, for example from “strongly agree” to “strongly disagree”. Most people who have participated in a marketing survey or an opinion poll are familiar with such questions. The intensity of sentiment can be later plugged into a predictive, or a descriptive model.

Similarly, for political data analysis, tracking just the polarity of opinion is seldom useful: changes from clearly positive to clearly negative, or vice versa, are rare, and seldom occur overnight.  On the other hand, the intensity of opinion changes constantly, and a trend can signal that the change of polarity is coming.  Such trending, cross-referenced with other data, can produce useful information for decision-making.

We feel that new sentiment analysis tools should be at least on par with their traditionally used predecessors. SentiMetrix’s patent-pending SentiGrade™ technology measures the intensity of opinions on a contiguous scale.  This scale can be calibrated to meet the needs of a specific customer and their research problem.