maytepenella
commented
6 years ago text.py includes two new functions:
- encode_sentiment() to assign a number to each sentiment (negative-0,neutral-1,positive-2)
- correlation_to_sentiment -- function for correlation. It computes correlation between encoded sentiment and features and lists them (all or N features) in descending order to absolute correlation value.
Positive correlations correspond to features more important in possitive tweets and negative correlations in negative tweets
Find which features are more correlated with sentiment. List all the features or some and order them by correlation (absolute value). Filter by correlation value?