MIT Press

Table 19

Emotion prediction performance of appraisal-aware text classifiers (⁠ ${T \to E}_{model}$ ⁠, ${{TA}^{Gold} \to E}_{model}$ ⁠, ${{TA}^{Pred} \to E}_{model}$ ⁠).

Emotion	$\binom{T \to E}{human}$	$\binom{T \to E}{model}$	$\binom{{TA}^{Gold} \to E}{model}$	$\binom{{TA}^{Pred} \to E}{model}$
Emotion	F₁	F₁	F₁	F₁
Anger	.57	.53_±.05	.57_±.02	.57_±.02
Boredom	.73	.84_±.01	.83_±.03	.83_±.03
Disgust	.65	.66_±.00	.66_±.04	.66_±.04
Fear	.73	.65_±.03	.67_±.04	.67_±.03
Guilt	.53	.48_±.06	.58_±.05	.56_±.07
Joy	.49	.45_±.02	.48_±.03	.47_±.03
No emotion	.33	.55_±.01	.56_±.02	.56_±.01
Pride	.59	.54_±.03	.55_±.01	.55_±.01
Relief	.64	.63_±.02	.62_±.01	.62_±.02
Sadness	.63	.59_±.03	.65_±.01	.63_±.00
Shame	.48	.51_±.01	.50_±.08	.49_±.07
Surprise	.42	.53_±.02	.49_±.03	.50_±.02
Trust	.52	.74_±.02	.73_±.04	.72_±.03

Macro avg.	.56	.59_±.01	.61_±.02	.60_±.02

Emotion	$\binom{T \to E}{human}$	$\binom{T \to E}{model}$	$\binom{{TA}^{Gold} \to E}{model}$	$\binom{{TA}^{Pred} \to E}{model}$
Emotion	F₁	F₁	F₁	F₁
Anger	.57	.53_±.05	.57_±.02	.57_±.02
Boredom	.73	.84_±.01	.83_±.03	.83_±.03
Disgust	.65	.66_±.00	.66_±.04	.66_±.04
Fear	.73	.65_±.03	.67_±.04	.67_±.03
Guilt	.53	.48_±.06	.58_±.05	.56_±.07
Joy	.49	.45_±.02	.48_±.03	.47_±.03
No emotion	.33	.55_±.01	.56_±.02	.56_±.01
Pride	.59	.54_±.03	.55_±.01	.55_±.01
Relief	.64	.63_±.02	.62_±.01	.62_±.02
Sadness	.63	.59_±.03	.65_±.01	.63_±.00
Shame	.48	.51_±.01	.50_±.08	.49_±.07
Surprise	.42	.53_±.02	.49_±.03	.50_±.02
Trust	.52	.74_±.02	.73_±.04	.72_±.03

Macro avg.	.56	.59_±.01	.61_±.02	.60_±.02