Table 5.

LOSO RMSE scores for each of the classifiers. The results for the best-performing models for each feature type are red.

Features	Dim. Red. (n_comp)	LR	DT	1NN	SVM	GradBoost
LIWC	None	10.067	5.766	5.626	6.083	$\text{4.014}$
	LDA (23)	8.928	8.738	5.224	6.195	7.654
	PCA (20)	4.436	5.383	5.364	6.057	4.640
BERT	None	5.111	5.984	$\text{4.953}$	6.111	5.407
	LDA (23)	5.111	6.571	5.805	6.275	6.701
	PCA (2)	6.304	5.628	5.851	6.187	6.034
BERT + LIWC	None	9.475	4.956	4.752	5.919	$\text{4.050}$
	LDA (23)	8.515	8.038	5.285	6.821	7.234
	PCA (20)	4.574	5.228	5.680	5.165	4.509
BERT + CLAN	None	4.810	6.265	4.728	6.009	4.100
	LDA (23)	4.810	5.700	4.988	6.173	5.447
	PCA (20)	3.991	5.459	4.842	5.254	$\text{3.969}$
BERT + LIWC + CLAN	None	13.877	5.533	4.420	5.846	4.190
	LDA (23)	5.243	5.398	5.482	6.477	5.031
	PCA (20)	$\text{3.774}$	5.701	5.023	4.966	4.201
word vectors	None	5.294	5.467	5.204	6.146	5.684
	LDA (23)	5.294	5.158	$\text{4.967}$	5.936	5.228
	PCA (2)	6.359	6.061	5.958	6.148	6.241
	PCA (70)	5.419	5.561	4.981	6.177	5.516
i-vectors (VoxCeleb)	None	6.323	6.477	6.612	6.444	6.461
	LDA (23)	6.323	6.366	6.384	6.279	6.443
	PCA (2)	6.576	6.431	6.361	6.290	6.421
	PCA (10)	6.412	6.507	6.524	6.265	$\text{6.264}$
i-vectors (Pitt)	None	6.545	6.850	6.239	6.281	6.513
	LDA (23)	6.545	6.524	6.307	6.244	6.499
	PCA (2)	6.624	6.606	6.484	6.323	6.598
	PCA (20)	6.523	6.575	6.577	$\text{6.207}$	6.511
i-vectors (VoxCeleb + Pitt)	None	6.298	6.363	6.545	6.243	6.445
	LDA (23)	6.298	6.399	$\text{6.110}$	6.231	6.459
	PCA (20)	6.502	6.558	6.655	6.256	6.475
x-vectors (VoxCeleb)	None	6.424	6.400	6.208	6.400	6.369
	LDA (23)	6.424	6.478	6.493	$\text{6.162}$	6.413
	PCA (2)	6.618	6.767	6.531	6.381	6.634
	PCA (40)	6.246	6.320	6.517	6.329	6.378
x-vectors (Pitt)	None	6.310	6.534	6.445	6.405	6.504
	LDA (23)	6.310	$\text{6.073}$	6.403	6.245	6.318
	PCA (40)	6.471	6.456	6.181	6.369	6.474
x-vectors (VoxCeleb + Pitt)	None	6.385	6.268	6.394	6.401	6.386
	LDA (23)	6.385	6.379	6.230	$\text{6.170}$	6.442
	PCA (40)	6.296	6.433	6.411	6.288	6.467