Event Date
"Prosodic Focus in Speech Directed Toward Human vs. Voice-AI Interlocutors"
Millions of people now talk to voice-activated artificially intelligent (voice-AI) assistants, like Amazon’s Alexa and Apple’s Siri in their everyday lives. Yet, we still don’t fully understand how people talk to voice-AI versus real humans. This project tests how people emphasize parts of their sentence through prosody, by increasing their pitch, loudness, and duration; for example, saying “JADE saw the sun” emphasizes “Jade” while “Jade saw the SUN” emphasizes “sun”. We compare speech directed toward a human and an Amazon Alexa voice. In a second experiment, we test whether a separate group of listeners can identify whether the speakers in the first experiment had been talking to a human or voice-AI interlocutor. Our results show that speakers use similar prosody to emphasize words for both human and voice-AI talkers; this suggests that speakers are applying speech behaviors from human-human interaction to voice-AI. Yet, there are still nuanced differences for the two types of interlocutors, which suggests that speakers can change the way they use prosody to emphasize words based on their listener.
For more information please contact Sophia Minnillo, smminnillo@ucdavis.edu
Event sponsored by the Cluster on Language Research