Voice User Interface Knowledge Acquisition via Collaborative Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice-user interface (VUI) systems often fail to provide accurate answers due to misunderstandings of user questions or lack of knowledge about the entities involved, as they rely on programmed databases and data mining without leveraging user contributions for knowledge augmentation.
Innovation Solution
A VUI knowledge acquisition framework that utilizes collaborative filtering and machine learning to identify knowledge gaps, crowdsources high-quality answers from users, and incentivizes participation through rewards and gamification, allowing the system to expand its knowledge base and validate new facts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If VUI systems rely on programmed databases and data mining for knowledge acquisition, then the system structure remains simple and controllable, but the knowledge base coverage and accuracy are insufficient
Solution Approach 1:
The patent introduces a crowdsourcing platform as an intermediary between the VUI system and external knowledge sources. This platform mediates the knowledge acquisition process by coordinating multiple human annotators, managing task distribution, and aggregating results, thereby enabling the VUI system to access broader knowledge without directly implementing complex knowledge gathering mechanisms itself
Solution Approach 2:
The system implements self-service through automated confidence scoring and validation mechanisms. The VUI system automatically evaluates the reliability of crowd-sourced answers using confidence scores derived from multiple annotators' agreements and statistical measures, reducing the need for manual verification while maintaining knowledge quality
2Productivity
If VUI systems use traditional data mining methods, then the implementation process is straightforward, but the knowledge acquisition speed and scalability are limited
Solution Approach 1:
The patent applies preliminary action by pre-defining structured annotation schemas, taxonomies, and guidelines before the crowdsourcing process begins. This preparation work establishes clear frameworks for annotators to follow, enabling rapid and consistent knowledge acquisition without requiring time-consuming ad-hoc annotation discussions or revisions
Solution Approach 2:
The system uses partial action by implementing confidence scoring thresholds that determine when enough annotator input has been gathered. Rather than requiring unanimous agreement or fixed numbers of annotators for all cases, the system stops annotation processes when confidence scores reach predetermined thresholds, reducing unnecessary annotation time while maintaining adequate knowledge quality
3Adaptability or versatility
If VUI systems lack user-contributed knowledge, then the system design remains simple, but the system cannot adapt to specialized domains and evolving fields
Solution Approach 1:
The patent implements universality by designing a multi-functional crowdsourcing platform that handles diverse knowledge acquisition tasks across multiple domains. The same infrastructure supports various annotation types, domain-specific taxonomies, and knowledge structures, allowing the VUI system to adapt to specialized fields like medicine and rapidly evolving domains without requiring separate specialized systems for each
Data Source
AI summary
A voice user interface (VUI) system use collaborative filtering to expand its own knowledge base. The system is designed to improve the accuracy and performance of the Natural Language Understanding (NLU) processing that underlies VUIs. The system leverages the knowledge of system users to crowdsource new information.