How to define intensity levels for twitter data for sentiment analysis

2020-03-26 machine-learning sentiment-analysis text-classification

I have sample data of specific twitter users, which may not be correlated with any of the existing labeled datasets with intensity levels.

So I want to define an intensity level (score) for each tweet from, let say -3 (for extreme sadness) to +3 (for extreme happiness).

The data after having defined intensity levels will be used for classification (e.g. Continuously sad users are depressed). Can anyone suggest how can I define different intensity levels for the tweets in such datasets, and what should be the criteria of assigning intensity levels. I mean should I follow depression symptoms and depression diagnostic criteria for searching of terms on the basis of which a tweet is assigned -3,-2 or -1 intensity level? Any directions towards how to begin it will be appreciated.