The first detection is to check each post for how much emotionality is detected.
For this we use an independentent, objective classifier that simply indicates the level of emotional text used, if at all.*
*This emotional pre-processing can be enabled / disabled when you upload your dataset.
Next each post gets a score for how positive or negative the text is (sentiment or valence) as well as the balence of arousal and calmness.
Polarity and arousal levels are the net value (difference) between positive/negative detections and arousal/calmness.
To go beyond the postive or negative context, the post is then scored for each of 6 'core emotions'.
These include Joy, Anger, Sadness, Excitement, Fear, and Empowerment.*
Pushing beyond 'core emotions', posts are also analyzed for 27 different fine-grain, specific feelings that fall under the core emotion families listed above.
For example, this 'angry' post used words that indicated language that more specifically relates to 'annoyance'.