In-Vehicle Speech Interaction for Cross-Domain Destination Inheritance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current in-vehicle navigation systems separate navigation and weather functions, leading to a poor user experience as users cannot seamlessly check weather at the destination during navigation without repeating the destination's physical location.
Innovation Solution
A speech interaction method that enables cross-domain inheritance by recognizing keywords in user requests, acquiring APIs from historical interactions, performing parameter replacement and slot identification, and predicting APIs for seamless integration of navigation and weather functions over multiple rounds of interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If navigation and weather functions are separated into independent systems, then each function can be processed independently with clear system boundaries, but users must repeat the destination location when checking weather during navigation, degrading user experience
Solution Approach 1:
The patent merges navigation and weather functions into a unified speech interaction system. The server integrates multiple domain processing capabilities (navigation, weather, etc.) into a single system that can handle cross-domain requests. When a user asks about weather at the destination, the system combines navigation context (current destination) with weather querying functionality, eliminating the need for users to repeat location information and improving ease of operation.
Solution Approach 2:
The speech interaction server is designed with multi-functionality to handle various domains including navigation, weather, and other services through a unified interface. The system can process single-domain requests and cross-domain requests alike, adapting its behavior based on the user's intent and available context. This universal design allows the system to serve multiple purposes without requiring separate specialized systems.
2Adaptability or versatility
If the system stores historical APIs for multiple domains, then cross-domain entity inheritance can be achieved over long time spans, but the system complexity and storage requirements increase
Solution Approach 1:
The system performs preliminary action by pre-storing historical APIs and their associated entity information in a structured format. During speech interactions, the system has already prepared the necessary API templates and entity associations in advance, allowing rapid retrieval and adaptation when cross-domain requests occur. This pre-processing reduces the computational burden during actual user interactions and enables efficient cross-domain entity inheritance.
Solution Approach 2:
The patent introduces an intermediary mechanism in the form of a unified speech interaction server that mediates between different domain-specific functions. This server acts as a mediator that receives user requests, retrieves relevant historical APIs from stored templates, performs entity inheritance across domains, and coordinates the response generation. The intermediary structure manages system complexity by centralizing the integration logic rather than requiring direct connections between all domain modules.
3Measurement precision
If the system processes speech requests across multiple domains with context inheritance, then accurate entity recognition is maintained over multiple interaction rounds, but the processing time and computational resources increase
Solution Approach 1:
The system extracts and stores key entity information and API templates from historical interactions in a structured format. When processing new speech requests, the system extracts only the necessary entities and parameters from the user's input rather than reprocessing the entire context. This extraction approach maintains entity recognition accuracy by focusing on relevant information while reducing overall processing time by avoiding redundant computations.
Solution Approach 2:
The patent uses copying by retrieving and reusing previously processed API templates and entity structures from historical interactions. Instead of rebuilding the entire processing pipeline for each new request, the system copies relevant API definitions and entity associations from stored historical data, adapts them to the current context, and uses them for rapid processing. This copying mechanism preserves accuracy by reusing proven templates while significantly reducing processing time.
Data Source
Figure 1~3
Figure 4~7
Figure 8~9
AI summary
Disclosed are a speech interaction method, a server and a storage medium. The method includes: receiving a user speech request related to a first domain forwarded by a vehicle, and recognizing a keyword in the speech request; acquiring, by searching with the keyword, an application programming interface, called API, for a second domain from a set of historical APIs for speech interactions; performing replacement of the keyword and slot identification on the speech request based on a parameter name and a parameter value in the acquired API to generate a new speech request; predicating an API for the new speech request; and performing API parameter filling on the predicted API based on the parameter name and the parameter value in the acquired API and a result of the slot identification, and outputting and issuing a filling result to the vehicle to complete a speech interaction. The method can realize cross-domain inheritance of an entity over a long time span, and can thus meet users' needs during navigation for checking of weather at a destination and the like at any time after multiple speech interactions with the vehicle.