Skip to main content

Table 3 Experimental results on the development set of the DAIC-WOZ data set

From: Towards automatic text-based estimation of depression through symptom prediction

Model

Binary Diagnosis Eval

PHQ-8 Score Severity Eval

5-Class Severity Eval

\(miF_1\) \(\pm \sigma\)

\(maF_1\) \(\pm \sigma\)

\(\text {MAE}\) \(\pm \sigma\)

\(ma\text {MAE}\) \(\pm \sigma\)

miF1-5c \(\pm \sigma\)

maF1-5c \(\pm \sigma\)

Binary Diagnosis

0.806 ± 0.031

0.798 ± 0.031

-

-

-

-

5-Class Diagnosis

0.739 ± 0.049

0.713 ± 0.058

-

-

0.503 ± 0.049

0.237 ± 0.017

PHQ-8 Score Diagnosis

0.600 ± 0.030

0.507 ± 0.026

5.51 ± 0.06

6.01 ± 0.08

0.255 ± 0.024

0.159 ± 0.018

Symptom-based Diagnosis

0.752 ± 0.035

0.719 ± 0.047

3.61 ± 0.12

4.11 ± 0.18

0.442 ± 0.106

0.286 ± 0.063

  1. All models were run five times with different seed values, and the average values with standard deviation are presented; miF1-5c (resp. maF1-5c) stands for the 5-class micro-averaged F1-score (resp. macro-averaged F1-score). Bold values indicates the best results for each model