Natural Language Query Translation for Database-Specific Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques limit users' ability to access data stored in software applications due to template-based dashboards and the requirement for database-specific syntax, restricting data retrieval capabilities and user knowledge.
Innovation Solution
A machine learning model is trained to translate natural language queries into database queries, utilizing a pre-trained 'teacher' model to enhance domain-specific accuracy and resource efficiency, allowing users to access data without needing database-specific syntax.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If template-based dashboards and reports are used for data access, then users can easily access data through predefined interfaces, but users are restricted in their ability to request data according to specific needs and cannot extract full value from stored data
Solution Approach 1:
The patent introduces a translation service as an intermediary component that converts natural language queries into database-specific syntax. This mediator layer allows users to interact with the database using simple natural language while the translation service handles the complexity of syntax conversion, thereby maintaining ease of operation while enabling versatile data requests without requiring users to learn database syntax
Solution Approach 2:
The system enables a single interface to serve multiple functions: it can handle both predefined template-based queries and free-form natural language queries. The translation service makes the system universal by accommodating different types of user needs (both simple and complex queries) through a unified natural language interface, eliminating the need for separate template-based and custom query interfaces
2Measurement precision
If database-specific syntax is required for data retrieval, then precise data queries can be executed, but many users lack the knowledge to use these syntax requirements effectively
Solution Approach 1:
The translation service acts as an intermediary that bridges the gap between user-friendly natural language and precise database syntax. Users input natural language queries without needing to know database syntax, and the translation service automatically converts these into precise database queries, thereby maintaining query precision while improving user accessibility
Solution Approach 2:
The system creates a copy or translation of the user's natural language intent into the target database syntax. Instead of requiring users to directly write database syntax, the system generates an equivalent query in the target language through the translation service, preserving the precision of the original intent while removing the syntax barrier
3Measurement precision
If a pre-trained teacher model is used to train a target machine learning model, then domain-specific accuracy and resource efficiency are enhanced, but additional training infrastructure and processes are required
Solution Approach 1:
The patent applies preliminary action by using a pre-trained teacher model that has already learned general language patterns and knowledge before being used to train the domain-specific target model. This preliminary training of the teacher model on general data prepares it to effectively guide the subsequent domain-specific training, achieving high accuracy while managing training complexity through staged preparation
4Measurement precision
If a large machine learning model is used for natural language to database query translation, then translation accuracy is improved, but resource consumption increases
Solution Approach 1:
The patent applies parameter changes by training a smaller target model with domain-specific parameters and knowledge transferred from the larger teacher model. This changes the model parameters from general-purpose large-scale parameters to domain-optimized parameters, achieving high translation accuracy in the specific domain while reducing computing resource consumption compared to using the full large model
Data Source
AI summary
The present disclosure provides an approach for training a machine learning model. Embodiments include receiving text comprising a natural language request. Embodiments include providing one or more inputs to a source machine learning model based on the text, wherein the source machine learning model has been trained using source training data corresponding to a plurality of databases. Embodiments include receiving, from the source machine learning model in response to the one or more inputs, a database query in a syntax corresponding to a target database. Embodiments include generating training data for a target machine learning model based on the text and the database query received from the source machine learning model, wherein the target machine learning model has been trained using a smaller amount of training data than the source training data that was used to train the source machine learning model.