What is the ultimate formula for calculation of sentiment score across multiple items?
Because sentiment scores for individual documents are highly dependent on the amount and nature of the content in each document, it is not efficient to apply pure math operations across sentiment scores. Averaging multiple sentiment scores can give a skewed overall picture. Instead, looking at aggregate counts and what they represent is what is recommended. We’ve developed a “P/N ratio” formula that gives average sentiment score across documents.
How to calculate a “P/N ratio” for the sentiment score?
The PN ratio calculation formula is aimed for getting an average sentiment for the group of the records with sentiment scores. Let’s say you have 100 records with the Sentiment score value for all of them. For first you need to calculate the count of positive and negative records based on their sentiment score. The formula in pseudo code is:
IF positive_count < negative_count THEN
ratio = -1*(negative_count/positive_count)
IF positive_count > negative_count THEN
ratio = positive_count/ negative_count
ELSE ratio = 0
IF ratio < -10 THEN ratio = -10
IF ratio > 10 THEN ratio = 10