Skip to main content
. Author manuscript; available in PMC: 2022 Apr 30.
Published in final edited form as: Interspeech. 2021 Aug-Sep;2021:3830–3834. doi: 10.21437/interspeech.2021-2002

Table 3:

Results of multimodal methods on training set LOSO cross-validation

FS Feature Set Accu. F1 Spec.
Early g-LIWC+eGeMAPS 81.93 82.56 82.28
g-BERTword+eGeMAPS 75.90 76.74 75.95
g-BERTsent+eGeMAPS 74.10 75.71 70.89

g-LIWC+ComParE 71.69 73.45 68.35
g-BERTword+ComParE 74.70 75.58 74.68
g-BERTsent+ComParE 75.30 76.02 75.95

g-LIWC+Paraling. 75.90 76.74 75.95
g-BERTword+Paraling. 77.11 77.91 77.22
g-BERTsent+Paraling. 78.92 80.23 75.95

Late g-LIWC + eGeMAPS 72.29 74.16 68.35
g-BERTword + eGeMAPS 77.71 78.61 77.22
g-BERTsent + eGeMAPS 75.90 77.01 74.68

g-LIWC + ComParE 71.08 72.09 70.89
g-BERTword + ComParE 72.89 74.29 70.89
g-BERTsent + ComParE 72.29 73.26 72.16

g-LIWC + Paraling. 73.49 74.42 73.42
g-BERTword + Paraling. 72.29 74.73 65.82
g-BERTsent + Paraling. 71.08 71.76 72.15

Ens. Overall Avg. 75.3 76.84 72.15
Overall Maj. 77.11 79.12 70.89
Overall Wgt. 80.12 80.92 79.75

Top-10 Avg. 79.52 80.23 79.75
Top-10 Maj. 80.72 81.18 82.28
Top-10 Wgt. 79.52 80.23 79.75