Marking prosodic prominence for voice assistant and human addressees

Area of Science:

• Psycholinguistics
• Speech Communication
• Human-Computer Interaction

Background:

• Prosodic prominence guides listener attention by emphasizing new information.
• Audience design theory suggests speakers adapt communication based on listener abilities.
• The rise of voice assistants necessitates understanding human language adaptation to technology.

Purpose of the Study:

• To investigate prosodic prominence production and perception when addressing humans versus voice assistants.
• To determine if speakers adjust prosodic prominence based on an interlocutor's perceived communicative ability (audience design).
• To examine how the language system adapts to human-like technology.

Main Methods:

• California English speakers read sentences in response to questions probing different foci.
• Acoustic analyses measured prosodic prominence features (intensity, duration, pitch).
• Listener perception experiments assessed the inference of information structure from acoustic cues.

Main Results:

• Speakers consistently used acoustic adjustments to mark focus on sentence subjects or objects.
• Listeners accurately inferred intended information structure from these acoustic cues.
• No significant difference in focus marking was found between human and voice assistant interlocutors, but slower speech rates were associated with voice assistant interactions.

Conclusions:

• Speakers apply human-human interaction strategies when addressing voice assistants.
• Listeners, however, expect and perceive a distinct register for voice assistant communication.
• Findings suggest a nuanced adaptation of language, with differing expectations from speakers and listeners in human-voice assistant interactions.