Researchers at the Weizmann Institute of Science have discovered that prosody, or the melody of speech, operates as a distinct language with its own vocabulary and syntax. This finding, published in PNAS, suggests that AI could leverage prosody to understand language beyond just the words being spoken.
Prosody encompasses variations in pitch, loudness, tempo, and sound quality, adding a nuanced layer of meaning. The study identified recurring melodic patterns in spontaneous conversations, revealing hundreds of basic patterns that function as prosodic "words," each conveying specific attitudes like curiosity or surprise.
The AI model identified syntactic rules governing the order of these prosodic patterns, which often appear in pairs, functioning as simple sentences. This breakthrough could enable AI to better interpret human emotion and intent, leading to more adaptive and natural responses in AI systems. According to Prof. Moses, this research could pave the way for automated systems that compile dictionaries of prosody for various languages, enhancing AI's ability to understand and respond to human speech.