The problem is simple: there's too much information out there and no accountability for those providing it.
PoliticsEQ is an attempt to harness technology to improve our political discourse and consumption. It may seem curious that tracking 'emotional' readings could somehow improve our consumption of media. This is just a first step and uses existing API's to analyze data. But the hope is to generate momentum for thinking about HOW we think about politics.
In the future we hope to use "Machine Learning" to provide all kinds of meta data about political communication: is it persuasive, is it descriptive, is it useful, what is the track record of the author, etc. The hope is that by making the SPEAKER or WRITER the focus of attention we can reward informative content and discourage junk.
Yes, the site currently looks like crap as it's just another Twitter Bootstrap site. The hope however is that the content will be compelling enough to attract investment and eventually provide a professionally developed UX. I currently provide all technical muscle needed to make this site work. If you would like to help or have any concerns please feel free to contact me:
Mike Van Winkle
This methodology is admittedly rough and subject to change. For every article analyzed, we collect keywords and each keyword has a sentiment estimate attached to it. Since publishers have different tones and approaches to news we calculate a handicap which is the aggregate keyword sentiment across all keywords for that publisher. We subtract that handicap from the aggregate keyword sentiment for "Trump" and "Clinton" keywords. This gives us an adjusted keyword aggregate sentiment. Then we calclate the distance between those values to gives us a bias value where the higher sentiment indicates the "direction".