83 percent of Americans are more comfortable discussing political issues online than
in person. Meanwhile, pollsters’ calls drop nine out of ten times. This disconnect highlights an
underlying issue; it’s extremely difficult to find out what US citizens actually believe.
Built from scratch in Python, FeedScore addresses this need. We’ve created an advanced polling suite that allows analysts, journalists, politicians and individuals, to survey millions of social media posts and accurately track public opinion surrounding their
issue. Slated for Spring 2018 release, FeedScore’s beta is already used by Gallup and BuzzFeed.
After a user submits a keyword, geographical location, sample size, and date range, FeedScore scours the web to find relevant posts. Using our proprietary machine learning technology, FeedScore then parses each post’s text to analyze the user’s stance, their approval regarding the issue. Unlike other text sentiment analysis programs on the market, we
have designed FeedScore to analyze approval, rather than positivity, and trained it specifically on real twitter data. This specific approach pays off; in side by side testing, we’re able to handily outperform both IBM Watson and Lexalytics, well-established technologies in the space, when analyzes twitter conversation. Social media is more than a beauty salon or high school, it’s the public forum of the digital age.
Last year, the LA Times’ poll, which exclusively predicted Trump’s election, based results on data collected online. As Americans, we have a right to accurate data, and traditional collection methods don’t cut it. FeedScore brings polling to the 21st century.