Tenant-Specific and Global Pretagging for Natural Language Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural language query systems in cloud platforms, particularly multi-tenant database systems, fail to accurately capture intent and nuance due to inadequate mapping techniques, leading to inadequate or inaccurate query results, and face challenges in data privacy and resource management.
Innovation Solution
Implementing a tenant-specific tagging model that identifies data type identifiers for a specific tenant and a global tagging model using named entity recognition to generate semantic search trees, ensuring data privacy by not using tenant-specific data for global model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a database administrator maintains a mapping of possible variations of natural language structures to valid database queries, then the system can support natural language queries, but the mapping techniques fail to capture intent or nuance resulting in inadequate or inaccurate query results
Solution Approach 1:
The patent replaces the mechanical mapping system (administrator-maintained lookup tables) with a machine learning-based semantic analysis system. The ML model analyzes the intent and nuance of natural language queries by learning from training data, substituting the rigid mechanical mapping approach with an adaptive intelligent system that can capture contextual meaning and generate accurate database queries.
Solution Approach 2:
The patent changes the parameters of the query processing system from static mapping rules to dynamic machine learning model parameters. The ML model's weights and biases are trained on natural language data to optimize query generation accuracy, allowing the system to adapt to different intents and nuances rather than relying on pre-defined mappings.
2Measurement precision
If tenant-specific data is used for global model training, then the global tagging model can be more accurate, but data privacy is compromised
Solution Approach 1:
The patent extracts only the necessary linguistic patterns and structural information from tenant-specific data for training the global model, while excluding sensitive proprietary information. The system separates useful training signals (language patterns, query structures) from sensitive data (business logic, proprietary terminology), extracting only what is needed for generalizing natural language understanding across tenants.
Solution Approach 2:
The patent introduces an intermediary processing layer that anonymizes and generalizes tenant-specific data before it is used for global model training. This intermediary process transforms sensitive tenant data into abstract linguistic patterns that preserve training value while removing privacy concerns, acting as a mediator between data utility and privacy protection.
3Measurement precision
If multiple tagging models are implemented for each tenant, then query accuracy is improved, but resource intensive configuration needs increase
Solution Approach 1:
The patent creates a universal global tagging model that serves multiple tenants simultaneously, eliminating the need for separate models for each tenant. The global model is trained on aggregated data from multiple tenants and can handle diverse query types across different tenants, providing multi-functionality that reduces system complexity while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary training of the global tagging model on aggregated tenant data before deployment. This preliminary action creates a pre-trained model that can be efficiently deployed to serve multiple tenants without requiring individual configuration or training for each tenant, reducing ongoing configuration complexity.
Data Source
AI summary
A database system may receive a natural language query that is associated with a tenant of a multi-tenant system. The natural language query may be parsed into a set of tokens, and the set of tokens may be tagged, using a tenant specific tagging model associated with the tenant, the set of tokens with at least one pre-configured data type identifier that is configured for the plurality of tenants. A global tagging model that supports the plurality of tenants of the multi-tenant system may tag the set of tokens with at least one category identifier. The global tagging model may use the natural language query and the pre-configured data type identifier to identify the at least one category identifier. The system may execute a database query on a database associated with the tenant using the at least one pre-configured data type identifier and the at least one category identifier.