This repo contains code for sentiment analysis using BERT, and a comparison with a naïve Bayes classifier (using a subjectivity filter) on two datasets: the first dataset was compiled by Andrew Maas et al. in the 2011 paper Learning Word Vectors for Sentiment Analysis, and the second dataset was introduced by Pang and Lee, and improved by Richard Socher et al. in the 2013 paper Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank. The method and results of this study can be read in the report Comparison of Models for Sentiment Classification of Movie Reviews available in the repo.
BERT notebook:
NB notebook: