1. Using Tweets to Analyze the 2016 Presidential Election.
Nicklaus Glyder, Mitesh Gurav, Jesse Hinson, Ashwin Kumar, Kunwar Deep Singh Tor, Ravi Teja Kandula
Sentiment Analysis of 2016 Election Using Twitter Data
Ashwin Kumar, Jesse Hinson, Nicklaus Glyder, Mitesh Gurav, Ravi Teja Kandula, Kunwar Deep Singh Toor
Goal: Perform sentiment analysis on Tweets collected during the third 2016 Presidential Debate (focused on Clinton and Trump).
Identify and visualize keyword content of this Twitter sentiment data.
Process Flow Diagram:
Use Kafka, Zookeper, and
Twitter API on Palmetto Cluster
Clean Data using NLTK
and Python
Mass Extraction of Tweets.
Sentiment Analysis
using VADER
Creation of Visualizations
using D3.js and AngularJS
Accuracy of VADER sentiment analysis:
We hard coded the sentiment classification of 107 tweets, and analyzed how VADER performed on this training set. By disregarding Tweets with weak
sentiment scores, we were able to achieve 68% accuracy in polarity classification. This shows how Tweets (especially those that are political) are hard to classify
correctly due to intrinsic context such as humor and sarcasm.
Hillary Clinton Donald Trump
Visit http://debate-visualization.herokuapp.com to interact with these and
more visualizations that represent our data!
What is Sentiment Analysis?
The objective of sentiment analysis is to determine the positive/negative emotions
(polarity) associated with given text. We also want to determine extremity of the
sentiment. In other words, how positiver or negative was the sentiment of a tweet?
We used the VADER social media sentiment analyzer, a context-aware analyzer
developed at Georgia Tech.
C.J. Hutto and Eric Gilbert, "VADER: A Parsimonious Rule-based Model for Sentiment Analysis of Social Media Text",
Eighth International Conference on Weblogs and Social Media, Ann Arbor, MI, June 2014.
Motivations
* We wanted to use Tweets as a barometer for how people felt about different issues
during the 2016 Presidential Campaign.
* Social media and micro-blogging like Twitter have become a very popular place to
discuss politics; we were able to collect over 100,000 Tweets.
* Led to some interesting results: "terrorism" was found to be the most negative
word in both cases while "campaign", "women", and "lie" were the most frequently
used words.