We were tasked with analyzing social and news media to discover what really mattered to voters this election. And using artificial intelligence, we created something completely new. ITV asked us to present our creation during their Trump vs. Clinton Election Night results show.
We built a bespoke cognitive platform that was able to read and analyze a vast amount of unstructured data: All the candidate’s speeches and debates, hundreds of social feeds, 15 million news articles, and over a billion social posts. We used Natural Language Classifiers to narrow our consideration down to posts that implied voting intent, giving us two supporter bases of more than ten million individuals across the US—and indicating that Trump would be the victor. We then used IBM Watson tools to determine what those supporters of Trump and Clinton were talking about, and which topics carried the most importance and relevance in regards to their voting intent.
During the results show, we were able affirm that although Clinton supporters were winning the online debate and swaying more voters with the ‘Culture Wars’ (LGBT, gender equality, civil rights, and the environment), the Trump camp was winning more voters with their position on the economy, employment, and terrorism. From the outset of this project, our data also showed Trump would win. During the event, our map indicating voting intention predicted four out of the five swing states including Florida, and the ultimately the final result. As we discovered, reading the intent of millions of vocal voters can provide a very accurate guide to voting patterns. Eagle.Ai is now being developed as a tool for determining election outcome, as well as the issues that sway.