Table 3 Description of the top 15 acoustic and linguistic features (rank-ordered by importance) retained for the combined machine learning modeling

Delta energy entropy (max)

Acoustic (Prosody)

Entropy of energy can be interpreted as a measure of abrupt changes.

Delta energy (mean)

Acoustic (Prosody)

Dynamic changes in energy relate to the perceptual characteristics of pitch and loudness.

Energy contour (kurtosis)

Acoustic (Prosody)

The kurtosis of the energy contour for voiced segments.

Delta F0 (std)

Acoustic (Prosody)

First derivative of the fundamental Frequency. Reduced F0 can indicate low pitch and a flatter voice.

MFCC5 (max)

Acoustic (MFCC)

5th MFCC coefficient. It can describe vocal tract changes in voice spectral energy.

Superlative adverbs

Linguistic (POS)

Use of superlative adverbs (e.g., biggest, hardest, etc.)

OQ (skewness)

Acoustic (Glottal)

OQ is a measurement of the glottal flow. It can differentiate between a breathy and tense voice.

Achievement language

Linguistic (LIWC)

Use of agentic language as defined by the LIWC achievement dictionary (words such as win, success, better)

Delta chroma 2 (min)

Acoustic (Prosody)

A representation of spectral energy. Chroma-based features are also referred to as “pitch class profiles“.

Proper nouns

Linguistic (POS)

Use of singular proper nouns.

Delta energy (median)

Acoustic (Prosody)

Dynamic changes in energy relate to the perceptual characteristics of pitch and loudness.

Delta chroma 1 (median)

Acoustic (Prosody)

A representation of spectral energy. Chroma-based features are also referred to as “pitch class profiles“.

Delta MFCC5 (max)

Acoustic (MFCC)

The frame-based delta of the 5th MFCC coefficient. It can measure vocal tract changes.

NAQ (mean)

Acoustic (Glottal)

Average NAQ is a measure of the glottal flow that can differentiate between a breathy and tense voice.

Possessive pronouns

Linguistic (POS)

Use of possessive pronouns (e.g., my, his, hers)