Skip to main content
. 2023 Apr 18;66(8 Suppl):3132–3150. doi: 10.1044/2023_JSLHR-22-00263

Figure 2.

A block diagram of the methodological process for the present study. Block 1: Spoken Dialogue. Block 2: Annotated for Individual Speaking Turns. 2 arrows labeled Interlocutor 1 and Interlocutor 2 are drawn from Block 1 to Block 2. Block 3 is for the Extraction of Acoustic Features and it has 5 subblocks labeled M F C C, L T A S, Voice Report, Rhythm Metrics, and E M S. An arrow is drawn from Block 2 to Block 3. Blocks 4 and 5 are for the Entrainment Score Calculation. Blocks 4 and 5 are labeled Proximity and Synchrony, respectively. Arrows are drawn from Block 3 to Block 4 and Block 5. Block 6 is for Statistical Analysis and it is labeled Predictive modeling, Elastic Net. Arrows are drawn from Blocks 4 and 5 to Block 6.

Overview of methodological process for this study. Spoken dialogs are divided into individual speaking turns. Moreover, 429 acoustic features (divided into five acoustic feature sets) are extracted from each speaking turn in every conversation. Proximity and synchrony scores are calculated for each acoustic feature, yielding 858 entrainment scores per speaking turn. Predictive modeling is used to evaluate the degree of entrainment (i.e., degree to which entrainment scores could be used to distinguish real and sham conversational turns) and the relationship between entrainment and conversational success (i.e., degree to which entrainment scores could be used to predict conversational efficiency and quality scores). EMS = envelope modulation spectrum; LTAS = long-term average spectrum; MFCC = mel-frequency cepstrum coefficient; VR = voice report.