Sentiment Analysis for Mass Media

The charts show computed correlation coefficients between sentiment scores for data consisting of mass media news titles. The sentiment scores are computed using a lightweight BERT (Bidirectional Encoder Representations from Transformers) a foundational pre-trained transformer model designed for a wide range of NLP tasks. An example being if a media is covering something positively like stating 'The weather is nice' then the media most likely gets a positive sentiment score meanwhile if the statement is negative like 'The weather is bad!' then a negative sentiment score is the result. Correlation between scores is then high when both are scored with positive sentiment. Use the visualizations below to analyze trends and performance.