Query Recognition System Dispatcher Context Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dialog systems struggle to accurately process user requests that are not understandable out of speech and environmental context, particularly when the number of commands is high, leading to limited supported commands per application and conflicts between commands.
Innovation Solution
A query recognition system comprising a dispatcher and multiple natural language processing entities, where the dispatcher prioritizes responses from recently activated entities and uses context rules to process user requests, allowing for the identification and prediction of missing context to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional dialog systems process each user request independently without context, then the system complexity is low, but the recognition accuracy deteriorates for requests that depend on speech or environmental context
Solution Approach 1:
The dialog system is segmented into multiple independent natural language processing entities, each specialized in recognizing specific types of user requests within particular contexts. This segmentation allows the system to maintain low overall complexity while improving recognition accuracy for context-dependent requests, as each entity handles only its specialized domain.
Solution Approach 2:
A dispatcher component is introduced as an intermediary that receives user requests and routes them to the appropriate natural language processing entity based on the request type and context. This mediator enables the system to handle complex context-dependent recognition tasks without requiring the entire system to be overly complex, as the dispatcher intelligently distributes work to specialized entities.
2Adaptability or versatility
If the dialog system includes all commands in one unified system, then the command recognition coverage is high, but command conflicts increase and system complexity worsens
Solution Approach 1:
The system divides the comprehensive command recognition task into multiple specialized natural language processing entities, each responsible for specific commands or command categories. This segmentation maintains high command coverage across the system while reducing conflicts within individual entities, as each entity handles a limited, well-defined subset of commands.
Solution Approach 2:
Each natural language processing entity is designed with local quality specialization, meaning each entity is optimized for recognizing specific types of commands or requests rather than attempting to recognize all commands. This specialization reduces command conflicts within each entity while the collective system maintains comprehensive command coverage through the dispatcher's routing capabilities.
3Productivity
If the dialog system processes user requests without considering speech and environmental context, then the processing speed is high, but the ability to understand context-dependent requests deteriorates
Solution Approach 1:
The natural language processing entities are pre-configured with knowledge of specific contexts (speech context, environmental context) relevant to their specialized domains. This preliminary preparation allows them to quickly process context-dependent requests without requiring complex real-time analysis, thereby maintaining high processing speed while improving context understanding capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosure relates to a query recognition system (100) for automatically recognizing a current linguistic user query, wherein the current linguistic user query is represented by digital values. The query recognition system (100) comprises: a group of natural language processing entities (131 a-c) for associating computer readable commands with linguistic user queries, the computer readable commands relating to specific computer services which are different for each language processing entity (131a-c); and a dispatcher (101) being configured to select a natural language processing entity (131a) from the group of natural language entities (131a-c) which most recently output a computer readable command, and to pass on the current linguistic user query to the selected natural language processing entity (131a); wherein the selected natural language processing entity (131 a) is configured to determine whether a computer readable command is associated with the current linguistic user query, and to output a computer readable command if the computer readable command is associated with the current linguistic user query.