List of good and bad words

[maytepenella]

Find out which words are more correlated to possitive and negative sentiment. Maybe we can use the matrix correlation to define best possitive and negative words (+ correlations are good and - correlations appear more in negative sentences)

[maytepenella]

Added function get_vocabulary_per_sentiment to train.py

• This funciton takes negative tweets, creates a count_vectorizer of bow_size2 words
• then takes possitive tweets, creates a count_vectorizer of bow_size2 words
• Prints correlations with sentiment
• Updates dictionary with possitive terms not found in negative tweets and creates a new countvectorizer with this new vocabulary.
• Final length of vocabulary is lower than 2* bow_size2 as there are always words that are present in both negative and possitve tweets.

TO DO: discard words with low correlation?

[maytepenella]

TO DO: Upload version with neutral vocabulary

[maytepenella]

Modified version that considers all the senitments present in df.

Code has been redone to be more understandable.

For now we will not discard words based on correlation

[maytepenella]

Change to suit with main.py:

• Consider language as input
• Check that works without warnings/errors

[maytepenella]

Fixed on last commit