Integrating Common and Context-Specific NLU Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing NLU frameworks struggle to adapt to context-specific virtual assistant applications, leading to inaccurate information, frustration for users, and inefficiencies in development due to limitations in concurrent editing, deployment, and maintaining performance across multiple teams and contexts.
Innovation Solution
A system that integrates common and context-specific natural language understanding processing by combining off-the-shelf NLU offerings with team-specific experiences, allowing for separation of NLU artifact training and response generation, enabling cross-team collaboration and independent operation while simplifying deployment and enhancing user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single NLU framework is used for multiple channels or contexts, then development effort is reduced, but response accuracy deteriorates
Solution Approach 1:
The patent segments the NLU framework into a common base framework and multiple context-specific extensions. Each development team works with the common framework while adding their own context-specific training data and configurations, allowing simultaneous development without conflicts and maintaining accuracy for each specific context.
Solution Approach 2:
The patent merges the common NLU framework with multiple context-specific extensions to create a unified system that serves multiple channels and contexts. The common framework provides shared functionality while context-specific additions ensure accurate responses for each particular use case.
2Productivity
If multiple development teams work on the same NLU framework, then resource utilization improves, but deployment coordination complexity increases
Solution Approach 1:
The patent divides the development workspace into separate namespaces for each team, allowing concurrent editing of the common NLU framework without conflicts. Each team can independently add context-specific training data and configurations without interfering with other teams' work, eliminating the need for complex coordination.
Solution Approach 2:
The patent creates isolated copies or namespaces for each development team within the common framework, allowing them to work independently on their context-specific extensions without affecting other teams. This copying approach enables parallel development while maintaining a unified common framework.
3Reliability
If context-specific training data is added to improve accuracy, then response quality improves, but system complexity increases
Solution Approach 1:
The patent merges context-specific training data with the common NLU framework in a unified structure. The context-specific data is integrated as extensions to the common framework rather than as separate standalone systems, maintaining response quality for each context while avoiding the complexity of managing multiple independent systems.
Data Source
AI summary
Methods and apparatuses are described for integrating common and context-specific natural language understanding (NLU) processing in a virtual assistant application. A server identifies a common NLU processing unit comprising (i) a common NLU classification model and (ii) a common response library comprising common responses. The server builds a combined NLU processing unit based upon the common NLU classification model and a context-specific NLU classification model, the combined NLU processing unit comprising (i) a combined NLU classification model trained using the common NLU training data and context-specific NLU training data, (ii) the common response library and (iii) a context-specific response library comprising context-specific responses. The server generates a response to an input utterance by determining a user intent associated with the input utterance using the combined NLU classification model, locating a response associated with the user intent in one or more of the libraries.