Voice Assistant Character Specialized Model Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice assistant services lack the ability to provide responses in a style specialized to specific characters, relying on rule-based algorithms rather than artificial intelligence, limiting their ability to adapt and personalize interactions.
Innovation Solution
A system and method for providing a voice assistant service using an artificial intelligence (AI) model specialized in a character, allowing for the reception, generation, and output of response messages based on a character-specific model, enabling personalized interactions through ASR, NLU, DM, NLG, and TTS processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If rule-based algorithms are used for voice assistant processing, then the system is simpler to implement and easier to control, but the adaptability and personalization capability are limited
Solution Approach 1:
The patent changes the fundamental parameter of the voice assistant system from rule-based algorithms to AI-based models. By introducing character specialized models that can be dynamically selected and applied, the system achieves adaptability without requiring complete system redesign, thus resolving the contradiction between adaptability and complexity.
Solution Approach 2:
The patent segments the voice assistant functionality into a base voice assistant program and separate character specialized models. This segmentation allows the core system to remain simple while enabling personalized character responses through modular AI models that can be independently selected and applied.
2Adaptability or versatility
If a single fixed style response is used for the voice assistant, then the system is easier to operate and maintain, but the personalization and user engagement are reduced
Solution Approach 1:
The patent introduces dynamic response style selection where the voice assistant can switch between different character specialized models based on user preferences or context. This dynamic capability enables personalization while maintaining ease of operation through automated model selection without requiring manual style configuration.
Solution Approach 2:
The system performs self-service by automatically selecting and applying appropriate character specialized models based on pre-configured user preferences or default settings. This eliminates the need for users to manually adjust response styles, maintaining ease of operation while achieving personalization through automated model selection.
3Measurement precision
If AI models are introduced for character specialized responses, then the personalization and natural language understanding improve, but the processing time and computational resources increase
Solution Approach 1:
The patent applies partial action by using AI models only for specific character specialized response generation rather than processing all voice assistant operations through heavy AI computation. The base voice assistant functions use lighter rule-based processing, while only the character-specific personalization aspects utilize computationally intensive AI models, thus balancing understanding quality with processing time.
Data Source
AI summary
Provided are an artificial intelligence (AI) system that utilizes a machine learning algorithm such as deep learning, etc., and an application of the AI system. A method performed by a device for providing a voice assistant service through a voice assistant program includes: receiving, from an external device, a character specialized model for the voice assistant program; receiving a user voice input including a request for a response of the voice assistant program and a word indicating a character; determining the character specialized model according to the word indicating the character; generating a response message to the request for the response of the voice assistant program, using the character specialized model; and outputting the generated response message.