Sentiment score formula for SENTIMENTAL ANALYSIS with NEUTRAL sentiments

Not all the sentiments are binary i.e., Positive and Negative. While performing sentimental analysis with lexicon approach, the sentimental score formula doesn’t mention anything about Neutral Sentiments. Thus, the overall sentimental prediction is influenced as it totally depends on the sentimental score.
I have planned to use the formula
(($Positive Words$ +$NEUTRAL Words$)- $Negative Words$) / $All Words$
instead of the one in the tutorial
($Positive Words$ - $Negative Words$) / $All Words$

Furthermore, I have tried changing the sentimental prediction condition as such
$Sentiment Score$ >= $${DMean(Sentiment Score)}$$ => “POS”
$Sentiment Score$ = 0 => “NEU”
However, I am not very sure whether these changes will address the issue of Neutral sentimental prediction using lexicon approach.

Hi @deepjg,

I assume you are referring to this workflow:*zp_hhUROHNXToZHX

While it is true that not every sentiment analysis is a binary classification into positive and negative, in this particular example we work with a prelabeled dataset. So this might indeed be a sample of movie reviews which consists of only positive and negative ratings. However, some of the reviews might be closer to neutral ratings than other and thus, introducing a third class (neutral) can be useful as well.

Thanks for sharing your thoughts on this.


Thank You for your reply @Marten_Pfannenschmidt
Yes, I am referring to the same workflow and was trying to perform sentimental analysis using the same method but was stuck when I had to also introduce the class ‘Neutral’, as I could not find any hint or tutorial on how to introduce this class. I am not sure whether modifying the Math formula for sentimental score should be adopted or changing the rule in Rule Engine will help. I am attaching the workflow with this reply. Kindly, suggest me your valuable suggestions on this issue. I shall be thankful to you.
WithNeutral.knwf (1.2 MB)