Oscars Predictor is a prediction model for the Academy Awards based on NLP of tweet text. Featured on FiveThirtyEight's 'Can the Internet Predict the Oscars?' competition.
Tweet data is collected via three twitter bots hosted on AWS Lambda
Over 100k tweets were collected and included in analysis.
Sentiment Analysis to determine the positivity of a given tweet was performed with nltk VADER. From the VADER repo:
VADER (Valence Aware Dictionary and sEntiment Reasoner) is a lexicon and rule-based sentiment analysis tool that is specifically attuned to sentiments expressed in social media.
Positively rated tweets that mention a movie or actor were counted as a 'vote'.
Predictions were generated by tallying votes in each tweet. Votes were then aggreagated in Pandas DatFrame and normalized to produce an expected winner.