Table 4 Results using the augmented versions of the data with short sentences removed.
From: Multimodal deep learning for dementia classification using text and audio
Shorts-augmented | Accuracy | Precision | Recall | F1-score | AUROC |
|---|---|---|---|---|---|
Audio | 0.5954 ± 0.001 | 0.5989 ± 0.025 | 0.1939 ± 0.029 | 0.2914 ± 0.03 | 0.6258 ± 0.008 |
Text | 0.841 ± 0.01 | 0.8212 ± 0.014 | 0.8089 ± 0.024 | 0.8148 ± 0.014 | 0.9276 ± 0.009 |
Audio \(+\) Time | 0.6254 ± 0.005 | 0.5894 ± 0.023 | 0.4298 ± 0.05 | 0.4951 ± 0.032 | 0.6692 ± 0.008 |
Text \(+\) Time | 0.8478 ± 0.003 | 0.8375 ± 0.02 | 0.8039 ± 0.207 | 0.8199 ± 0.007 | 0.9345 ± 0.005 |
Audio \(+\) Text | 0.835 ± 0.02 | 0.8154 ± 0.036 | 0.7982 ± 0.039 | 0.80591 ± 0.023 | 0.9216 ± 0.02 |
Audio \(+\) Text \(+\) Time | 0.8451 ± 0.003 | 0.83646 ± 0.027 | 0.7992 ± 0.028 | 0.8166 ± 0.04 | 0.931 ± 0.005 |
