Bugs in calculating sentiment

sloria / TextBlob

Simple, Pythonic, text processing--Sentiment analysis, part-of-speech tagging, noun phrase extraction, translation, and more.

https://textblob.readthedocs.io/

MIT License

9.11k stars 1.13k forks source link

Open pienio7 opened 3 years ago

pienio7 commented 3 years ago

I found bugs in calculating sentiment:

Sentiment for the phrase "definitely happy" is the same as sentiment for word "happy" (polarity=0.8, subjectivity=1.0) - but "definitely" has intensity="2.0", so polarity of "definitely happy" should by 1.0.
Sentiment for "very no happy" is (polarity=-0.5, subjectivity=1.0) and for "very not happy" is (polarity=-0.1, subjectivity=0.65) - but it should be the same.