Fig. 1: Paradigm design. | Communications Psychology

Fig. 1: Paradigm design.

From: Question-based computational language approach outperforms rating scales in quantifying emotional states

Fig. 1

The figure shows the paradigm used in the study. In phase 1, on the left-hand side, participants are instructed to write an autobiographical narrative of an episode where they experienced either depression, anxiety, satisfaction, or harmony. Following this, they summarize the emotions in the narrative with five descriptive words and fills in rating scales corresponding to the four emotions (i.e., PHQ-9, GAD-7, SWILS, and HILS respectively). The five words are quantified by a large language model (BERT) into a vector. A machine learning algorithm (multiple logistic regression) is used to categorize either the vector or the total score of the rating scales. This categorization is compared with the emotion that participants were instructed to use while writing the narrative. Phase 2 is identical to Phase 1, with the exception that the participants read the stories generated in Phase 1 (rather than writing them).

Back to article page