Table 6.

RMSE scores for classifiers evaluated on the test set. The results for the best-performing models during cross-validation are red.

Features	Dim. Red. (n_comp)	LR	DT	1NN	SVM	GradBoost
LIWC	None	36.974	7.303	6.403	6.465	$\text{4.862}$
	LDA (23)	12.286	9.657	7.388	6.313	8.365
	PCA (20)	4.422	5.967	5.990	6.431	4.383
BERT	None	5.365	5.640	$\text{4.923}$	6.169	4.883
	LDA (23)	5.365	7.515	6.017	6.253	7.373
	PCA (2)	5.661	5.858	6.287	6.067	5.691
BERT + LIWC	None	34.420	7.127	5.021	6.103	$\text{5.037}$
	LDA (23)	14.905	8.624	5.742	7.189	6.561
	PCA (20)	4.872	7.078	5.159	4.895	4.404
BERT + CLAN	None	4.991	7.218	4.515	6.097	4.901
	LDA (23)	4.991	6.523	5.600	6.422	6.660
	PCA (20)	4.764	7.577	6.413	5.218	$\text{4.560}$
BERT + LIWC + CLAN	None	15.465	6.112	4.811	6.023	4.724
	LDA (23)	8.110	6.500	5.753	6.887	6.021
	PCA (20)	$\text{4.800}$	6.196	5.532	4.794	5.087
word vectors	None	4.714	5.280	5.129	6.147	5.361
	LDA (23)	4.714	5.111	$\text{5.344}$	6.063	4.955
	PCA (2)	5.732	6.452	5.992	6.129	5.803
	PCA (70)	4.785	5.700	5.237	6.169	5.271
i-vectors (VoxCeleb)	None	6.600	6.305	6.269	6.161	6.396
	LDA (23)	6.600	7.056	6.360	6.461	6.820
	PCA (2)	6.194	6.514	6.546	5.999	6.237
	PCA (10)	6.335	6.840	6.298	6.110	$\text{6.386}$
i-vectors (Pitt)	None	6.530	6.622	6.758	6.142	6.170
	LDA (23)	6.530	6.712	6.133	5.956	6.473
	PCA (2)	6.225	6.827	6.370	6.151	6.342
	PCA (20)	6.257	6.278	6.110	$\text{6.199}$	6.252
i-vectors (VoxCeleb + Pitt)	None	6.292	6.042	7.391	6.158	6.145
	LDA (23)	6.292	6.567	$\text{5.694}$	5.905	6.407
	PCA (20)	6.316	6.439	6.607	6.168	6.431
x-vectors (VoxCeleb)	None	6.559	6.665	6.401	6.094	6.309
	LDA (23)	6.559	6.289	6.261	$\text{6.085}$	6.312
	PCA (2)	6.167	6.669	6.566	6.089	6.164
	PCA (40)	6.358	6.058	6.189	6.115	6.160
x-vectors (Pitt)	None	6.428	6.483	6.563	6.287	6.333
	LDA (23)	6.428	$\text{6.462}$	6.314	6.097	6.423
	PCA (40)	6.424	6.506	6.499	6.322	6.370
x-vectors (VoxCeleb + Pitt)	None	6.644	6.622	6.338	6.096	6.208
	LDA (23)	6.644	6.450	6.188	$\text{6.059}$	6.466
	PCA (40)	6.173	6.640	6.488	6.123	6.204