Measuring Entrainment in Spontaneous Code-switched Speech

TL;DR

This work studies code-switched spontaneous speech between humans, finding that patterns of written and spoken entrainment in monolingual settings largely generalize to code-switched settings, and some patterns of entrainment on code-switching in dialogue agent-generated text generalize to spontaneous code-switched speech.

Abstract

It is well-known that speakers who entrain to one another have more successful conversations than those who do not. Previous research has shown that interlocutors entrain on linguistic features in both written and spoken monolingual domains. More recent work on code-switched communication has also shown preliminary evidence of entrainment on certain aspects of code-switching (CSW). However, such studies of entrainment in code-switched domains have been extremely few and restricted to human-machine textual interactions. Our work studies code-switched spontaneous speech between humans, finding that (1) patterns of written and spoken entrainment in monolingual settings largely generalize to code-switched settings, and (2) some patterns of entrainment on code-switching in dialogue agent-generated text generalize to spontaneous code-switched speech. Our findings give rise to important implications for the potentially "universal" nature of entrainment as a communication phenomenon, and potential applications in inclusive and interactive speech technology.

Figures (3)

• Figure 1: The novelty of our work comes from incorporating acoustic-prosodic features in the study of entrainment in code-switched speech. Here we highlight the value of identifying multiple dimensions and feature sets of entrainment in CSW. While entrainment is evident by inspection of the lexical and CSW strategy features of the $S_{A1}$, $S_{B1}$ interaction (top) (blue = entrainment on a frequent word; green = entrainment on alternational CSW), the acoustic-prosodic features are essential for correctly identifying entrainment in the $S_{A2}$, $S_{B2}$ interaction (bottom) (yellow = acoustic-prosodic entrainment; pink = entrainment on CSW amount; purple, gray = lack of entrainment on CSW strategy).
• Figure 2: Proximity at the turn-level. Significant acoustic-prosodic features are indicated by * and dark orange bars. See Table \ref{['tab:proximity-turn']} in Appendix \ref{['sec:appendix']} for t and p-values corresponding to each feature.
• Figure 3: Convergence at the turn-level. All acoustic-prosodic features are significant. The percentage of weakly diverging conversations for each acoustic-prosodic feature is the difference between the percentage of weakly converging conversations and 100.