Multi-phrase Voice Response Generation for Natural Conversation Flow
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional chatbots in full duplex voice conversations typically provide single responses, which limits their interaction to 1:1 parity, failing to replicate the natural, bidirectional communication seen in human-to-human conversations where one party may speak more while the other listens and responds with supplementary thoughts.
Innovation Solution
Implementing a multi-phrase response mechanism that generates a primary response and supplementary responses, allowing for multiple voice outputs, including padding words or follow-up responses, to enhance the conversation flow and user experience by providing additional information during idle times or after primary responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single response is provided in full duplex voice conversation, then the system complexity is low, but the conversation naturalness and user engagement deteriorate
Solution Approach 1:
The patent segments the response into multiple independent voice outputs: a primary voice output containing the main response content, and one or more supplementary voice outputs containing additional information or follow-up thoughts. This segmentation allows the chatbot to replicate natural human conversation patterns where speakers often add supplementary comments or follow-up thoughts, thereby improving conversation naturalness without requiring complete system redesign
Solution Approach 2:
The system performs preliminary generation of both primary and supplementary responses in advance before the conversation continues. By pre-generating multiple potential voice outputs and their corresponding text representations, the system prepares multiple response options that can be selectively delivered, improving conversation flow and naturalness while managing complexity through advance preparation
2Ease of operation
If multiple voice outputs are generated and provided, then the conversation naturalness improves, but the processing time and system complexity increase
Solution Approach 1:
The system generates both primary and supplementary text representations in advance during the initial processing phase, before voice conversion. By preparing multiple response texts simultaneously rather than sequentially, and converting them to voice outputs in parallel where possible, the system reduces the total processing time penalty associated with generating multiple voice outputs
Solution Approach 2:
The system applies different processing qualities to different response components: the primary voice output receives full processing attention for accuracy and quality, while supplementary voice outputs may use optimized or streamlined processing paths. This local quality differentiation allows the system to maintain high quality for critical responses while reducing processing overhead for supplementary content, thereby managing overall processing time
3Ease of operation
If supplementary responses are provided during idle time, then the user engagement improves, but the information delivery efficiency may deteriorate
Solution Approach 1:
The system dynamically determines whether to deliver supplementary voice outputs based on real-time conversation state, user responses, and timing conditions. Rather than always providing supplementary responses, the system adaptively selects when supplementary outputs enhance engagement versus when they might delay critical information delivery, thereby balancing user engagement improvement with information delivery efficiency through dynamic decision-making
Data Source
Figure 1~2
Figure 3
Figure 4~6
AI summary
The present disclosure provides method and apparatus for responding in a voice conversation by an electronic conversational agent. A voice input may be received in an audio upstream. In response to the voice input, a primary response and at least one supplementary response may be generated. A primary voice output may be generated based on the primary response. At least one supplementary voice output may be generated based on the at least one supplementary response. The primary voice output and the at least one supplementary voice output may be provided in an audio downstream, wherein the at least one supplementary voice output is provided during a time period adjacent to the primary voice output in the audio downstream.