An example of using NLTK VADAR sentiment analyser to perform sentiment analysis on a Pandas dataframe.

vadar_sentiment.py

from nltk.sentiment.vader import SentimentIntensityAnalyzer
sent_i = SentimentIntensityAnalyzer()

def vadar_sentiment(text):
    """ Calculate and return the nltk vadar (lexicon method) sentiment """
    return sent_i.polarity_scores(text)['compound']

# create new column for vadar compound sentiment score
news_sentiments['vadar compound'] = news_sentiments['title'].apply(vadar_sentiment)

def categorise_sentiment(sentiment, neg_threshold=-0.05, pos_threshold=0.05):
    """ categorise the sentiment value as positive (1), negative (-1) 
        or neutral (0) based on given thresholds """
    if sentiment < neg_threshold:
        label = 'negative'
    elif sentiment > pos_threshold:
        label = 'positive'
    else:
        label = 'neutral'
    return label

# new col with vadar sentiment label based on vadar compound score
news_sentiments['vadar sentiment'] = news_sentiments['vadar compound'].apply(categorise_sentiment)

# plot distribution of article sentiments for each newspaper
plt.figure(figsize=(8,8))
sns.countplot(x='newspaper', data=news_sentiments, hue='vadar sentiment')
plt.title('VADAR Sentiment Analysis', weight='bold')
plt.ylabel('Number of Articles', fontsize=10, weight='bold')
plt.xlabel('VADAR Sentiment', fontsize=10, weight='bold')
plt.legend(bbox_to_anchor=(1.02, 1.0), loc=2, borderaxespad=0.0)
plt.show()