Can Sentiment Analysis Platform Predict Outcome of US Election?

Social media systems like Facebook and Twitter contain a wealth of information about how their users feel about current topics of interest — the trick is to be able to root through vast amounts of data and extract the hidden gems of information.

The word “sentiment” is defined as a view of, or attitude toward, a situation or event; an opinion; or a feeling of emotion. ZimGo, from BPU Holdings (BPU), is a Sentiment Analysis Platform that is designed to evaluate huge amounts of social media data to determine what people feel about particular topics.

ZimGo features a sophisticated pattern recognition engine. In the case of its natural language processing (NLP) capabilities, ZimGo augments regular dictionaries with the use of the Urban Dictionary, thereby allowing it to parse and understand posts using slang. Furthermore, ZimGo also includes emojis and hashtags in its analysis.

Of course, one topic of great interest is the US election, which is in progress at the time of this writing. The creators of ZimGo recently ran a sentiment analysis on the approval ratings for Donald Trump and Hilary Clinton associated with various key topics. The ensuing analysis can be presented in a variety of ways, including day-by-day and hour-by-hour (it's interesting to see how approval ratings varied by topic throughout the course of the presidential debates), but an overall summary is shown below.


Approval ratings for Trump vs. Clinton (Source: BPU Holdings)

Measuring areas like the economy, foreign affairs, healthcare, immigration, income distribution, and Washington culture, the ZimGo analysis revealed the public's close positive/negative feelings toward the candidates. The most recent debate only minimally affected these sentiments, perhaps confirming the level of follower commitment late in the election process. The most notable exception is that the sentiment toward Clinton is significantly more positive with regard to income distribution.

BPU Holdings recently announced plans to offer free licenses for the ZimGo sentiment analysis platform to university development partners in Korea and the US. Initially available in Korean, English and Japanese versions of ZimGo will soon be available in limited beta release. Those interested in joining the beta may enroll at ZimGo.cloud.

6 thoughts on “Can Sentiment Analysis Platform Predict Outcome of US Election?

  1. “The election results show a sharp divide between rural and urban voters. If social networks are biased to urban populations the results tend to exclude a large proportion of the voter base. Based on Tuesday's results this appears to explain failure of po

    Log in to Reply
  2. “The rural vs urban divide has been known about since at least 2000 so it should be taken into account by now. There may also be a divide between younger and older although that's much less pronounced.nThose most active on social networks tend to be young

    Log in to Reply
  3. “thank you for the article … As with many platforms, the depth of the data is limited only by the amount of resources we want to dedicate. The beauty of twitter is that it does include the GEO and the Demographic data with the posts, as well as the tech

    Log in to Reply
  4. “But as the USC/L.A. Times poll showed (since it was closer than any other major poll to being right) the correct weighting of the data is key to getting an accurate result. According to the article below, if you take their data and use the same weighting

    Log in to Reply

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.