Context-Aware Routing for Ambiguous Voice Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech processing systems face challenges in accurately determining the appropriate actions to perform or applications to use for ambiguous spoken utterances due to limited access to relevant contextual data during initial utterance meaning determination, leading to incorrect routing and inability to integrate new actions or data without modifying underlying natural language understanding (NLU) processing.
Innovation Solution
A context-aware routing system that segregates speech processing into domains and subdomains, using intra-domain routing with confidence providers and a contextual data management system to assess and manage contextual information for routing decisions, allowing for dynamic and accurate routing of utterances to the appropriate domains and applications based on contextual data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech processing systems use limited contextual data during initial utterance meaning determination, then processing speed is maintained, but routing accuracy deteriorates leading to incorrect action selection
Solution Approach 1:
The system performs preliminary contextual data gathering and analysis before final routing decisions are made. Contextual data is collected and prepared in advance through the contextual data management system, allowing the routing process to access pre-processed information quickly when an utterance occurs, thus improving accuracy without adding significant latency to the core processing path.
Solution Approach 2:
The speech processing system is segmented into multiple independent components: initial utterance processing, contextual data management, confidence provider evaluation, and routing decision modules. This segmentation allows the system to process utterances in parallel streams, where contextual data can be gathered and evaluated independently from the primary speech recognition path, reducing overall processing time while improving routing accuracy.
2Adaptability or versatility
If the system integrates new actions or contextual data sources, then system versatility improves, but NLU processing complexity increases requiring modifications to underlying processing
Solution Approach 1:
The system extracts contextual data management and routing decision logic from the core NLU processing pipeline into separate, independent modules. This extraction allows new actions and data sources to be integrated through the contextual data management system without requiring modifications to the underlying NLU processing, thereby improving extensibility while maintaining NLU simplicity.
Solution Approach 2:
The contextual data management system serves as a universal interface that handles multiple functions: collecting contextual data from various sources, evaluating relevance through confidence providers, and providing routing recommendations. This multi-functional component allows the system to accommodate new actions and data sources through a single standardized interface, avoiding the need to modify NLU processing for each new capability.
3Reliability
If the system uses comprehensive contextual data for routing decisions, then accuracy of interaction goals increases, but system complexity increases requiring multiple confidence providers and data management components
Solution Approach 1:
The system introduces a contextual data management system as an intermediary layer between raw contextual data sources and the routing decision process. This intermediary consolidates multiple data sources and confidence providers into a unified interface, allowing the system to leverage comprehensive contextual information for improved reliability while hiding the underlying complexity from the core routing logic.
Data Source
AI summary
A multi-tier domain is provided for processing user voice queries and making routing decisions for generating responses, including for user voice queries that include multi-domain trigger words or phrases. When an utterance is recognized as different intents in different domains, a routing system for a domain may consider contextual signals, including those associated with other domains, to determine whether the domain is the proper one to handle the request. This determination can be performed with a statistical model specifically trained to make such determinations using the available contextual data.