How does the ObamaMeter technology work?
The ObamaMeter technology first identifies every reference to Barack Obama in the media analyzed. It recognizes both direct references like "Barack Obama," and indirect ones such as "The President of the United States." Then, in a following pass, it identifies words that have a positive or negative connotation and are also grammatically connected to the reference to Obama - for example, "Obama's success," or "The president failed to realize..." We then output all of the references and all of the positive or negative language - thousands a day - and aggregate them into a score - how positive or negative the media are about Obama on a given day.
What is the underlying text analysis system for the ObamaMeter?
The core system is called Profiler Plus and is licensed from Social Science Automation, our parent company. The program is implemented in Common Lisp and, in contrast to most statistically-based Natural Language Processing systems, Profiler Plus is entirely rule-based providing fine grained control and complete process transparency. Profiler Plus performs text analytics in a highly flexible, multiple pass process. Results are then interpreted by our expert analysts and/or imported into such data analysis programs as Microsoft Excel, SPSS and SAS.
The Obamameter is available here.
I want to thank Hall Roger for this.