NLU Intent Transfer Validation Using Knowledge Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing domain-specific natural language understanding (NLU) systems require extensive retraining and data annotation to handle new tasks outside their original domain, making it time-consuming and expensive to expand their functionality.
Innovation Solution
The technique involves transferring NLU objects such as intents and entities between domains after validating their compatibility, using knowledge graphs and search logs to ensure seamless integration, allowing existing training data to recognize new objects without further training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual redesign of semantic schema and retraining of NLU models is performed to handle new tasks, then the system can cover new functions, but the process becomes time-consuming and expensive
Solution Approach 1:
The patent applies preliminary action by pre-validating intent transferability using knowledge graphs and search logs before actual domain deployment. The system pre-computes compatibility scores between intents and domains, so when new tasks need to be handled, the validation and transfer process is already prepared, eliminating the need for time-consuming manual semantic schema redesign and model retraining.
Solution Approach 2:
The system implements self-service by automatically validating intent transferability through knowledge graphs and search logs without requiring manual intervention. The validation process autonomously determines whether an intent from one domain can be transferred to another domain by checking compatibility criteria, enabling the system to expand its coverage independently without external retraining efforts.
2Adaptability or versatility
If manual redesign of semantic schema and retraining of NLU models is performed to handle new tasks, then the system can cover new functions, but the cost increases
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing intent transferability information in knowledge graphs and search logs. This preliminary preparation enables rapid, low-cost validation when new tasks are needed, avoiding the expensive process of manual semantic schema redesign and model retraining that would otherwise be required to expand system coverage.
Solution Approach 2:
The system implements self-service by autonomously validating intent transferability through automated processes using knowledge graphs and search logs. This eliminates the need for expensive manual intervention, data annotation, and model retraining, allowing the system to expand its functional coverage at minimal cost.
3Productivity
If NLU objects are transferred between domains without validation, then the expansion process is faster, but the reliability of the transferred objects decreases
Solution Approach 1:
The patent applies preliminary action by performing validation of intent transferability before actual domain transfer using knowledge graphs and search logs. This pre-validation ensures that only compatible intents are transferred, maintaining high reliability while enabling rapid expansion since the validation is already completed before deployment.
Solution Approach 2:
The system implements feedback by using search logs to validate whether transferred intents actually work in the target domain. The validation process provides feedback on transfer compatibility by analyzing whether the intent-phrases and slot-entity pairs from the source domain are appropriate for the target domain, ensuring reliable transfers while maintaining fast expansion through automated validation.
Data Source
AI summary
Aspects of the present invention provide a technique to validate the transfer of intents or entities between existing natural language model domains (hereafter “domain” or “NLU”) using click logs, a knowledge graph, or both. At least two different types of transfers are possible. Intents from a first domain may be transferred to a second domain. Alternatively or additionally, entities from the second domain may be transferred to an existing intent in the first domain. Either way, additional intent/entity pairs can be generated and validated. Before the new intent/entity pair is added to a domain, aspects of the present invention validate that the intent or entity is transferable between domains. Validation techniques that are consistent with aspects of the invention can use a knowledge graph, search query click logs, or both to validate a transfer of intents or entities from one domain to another.