So I'm performing sentiment analysis and testing different libraries against a labeled dataset. The best way seems to be to look at the sentiment labels (positive, negative, neutral) versus the raw polarity score because different packages seem to measure it differently (some produce on scale from -1 to 1, while some others return a 0, 1, or -1 and others simply return the sentiment label and no polarity score).
However, some packages simply return the polarity score on a scale from -1 to 1 without a sentiment label. I understand that 1 is positive, -1 is negative, and 0 is neutral but how do I classify numbers in between such as 0.17 or 0.75? What is the standard "threshold" cutoff (if any) for each? Is it simply anything > 0 is positive and anything < 0 is negative, and only 0.0 is neutral? Or is there a better way to label the sentiment?
Specifically, I'm looking at this issue using the TextBlob package.