The relationship between sentiment score and subjectivity

[ShiJunyiSJ1]

Could someone provide some insights or details regarding the methodology used to compute the subjectivity value in TextBlob? I'm particularly interested in understanding how this value is derived and its correlation, if any, with the sentiment score. Thank you！

[Vidito]

Could someone provide some insights or details regarding the methodology used to compute the subjectivity value in TextBlob? I'm particularly interested in understanding how this value is derived and its correlation, if any, with the sentiment score. Thank you！

TextBlob’s subjectivity comes from the Pattern library’s lexicon. Each word has a predefined subjectivity score (0 = objective, 1 = subjective).

For a sentence, TextBlob:

1. Looks up each word’s subjectivity
2. Applies simple modifiers (e.g., “very”, “not”)
3. Computes a weighted average of those values

So, subjectivity is essentially the average subjectivity of the words in the sentence.

---

Relationship to sentiment (polarity)

• Both are computed from the same lexicon, but represent different things:
  • Polarity → positive vs negative
  • Subjectivity → opinion vs fact

They often correlate (strong opinions tend to be subjective), but one does not directly determine the other.