Humane Speech Synthesis through Zero-Shot Emotion and Disfluency Generation

TL;DR

This work addresses the deficiency of emotional depth and disfluency in contemporary conversational AI by introducing a zero-shot emotion and disfluency generation framework. It leverages GPT-4 to embed emotion and disfluency cues in response text and uses a rule-based TTS mapping to realize these cues in speech, augmented by memory components to maintain context. The approach is evaluated in a Virtual Patient SBIRT training scenario, comparing neutral, moderate, and extreme prompts across multiple TTS models, revealing improvements in naturalness and perceived humanity of speech. The findings suggest that integrating prompt-driven emotional/disfluent cues with memory-aware dialogue and a rule-based speech synthesis pipeline can produce highly natural, human-like interactions with potential applications in empathetic healthcare training and beyond.

Abstract

Contemporary conversational systems often present a significant limitation: their responses lack the emotional depth and disfluent characteristic of human interactions. This absence becomes particularly noticeable when users seek more personalized and empathetic interactions. Consequently, this makes them seem mechanical and less relatable to human users. Recognizing this gap, we embarked on a journey to humanize machine communication, to ensure AI systems not only comprehend but also resonate. To address this shortcoming, we have designed an innovative speech synthesis pipeline. Within this framework, a cutting-edge language model introduces both human-like emotion and disfluencies in a zero-shot setting. These intricacies are seamlessly integrated into the generated text by the language model during text generation, allowing the system to mirror human speech patterns better, promoting more intuitive and natural user interactions. These generated elements are then adeptly transformed into corresponding speech patterns and emotive sounds using a rule-based approach during the text-to-speech phase. Based on our experiments, our novel system produces synthesized speech that's almost indistinguishable from genuine human communication, making each interaction feel more personal and authentic.

Figures (1)

• Figure 1: Proposed approach for emotion and disfluency generation in speech