LLM Query Translation Training With User Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Thousands of users are unable to perform or incorrectly execute data operations in complex programming languages like SPL due to their complexity and syntax requirements, limiting their ability to interact with databases effectively.
Innovation Solution
A system and method that utilizes generative artificial intelligence to translate natural language descriptions of search queries into executable code by receiving user feedback on multiple LLM translations, allowing users to generate training data for machine learning models like LLM 180, which improves translation accuracy over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If users directly write computer software code in complex programming languages like SPL, then precise data operations can be performed, but the complexity and syntax requirements make it inaccessible to most users
Solution Approach 1:
The patent introduces an intermediary system consisting of a machine learning model and natural language processing interface that translates between natural language and programming code. Users interact through natural language descriptions of desired data operations, and the system automatically generates the corresponding SPL code, eliminating the need for users to learn complex syntax while maintaining precise data operations.
Solution Approach 2:
The patent replaces the mechanical process of manually writing and debugging code with an automated machine learning-based code generation system. The system learns from training data to automatically generate correct SPL code from natural language inputs, substituting the manual coding process with an intelligent automated system that handles syntax and translation complexities.
2Reliability
If programming languages are designed to be complex with strict syntax requirements, then precise control over data operations is achieved, but thousands of users are unable to perform desired operations correctly
Solution Approach 1:
The system acts as an intermediary that ensures correctness by translating natural language into syntactically correct SPL code through machine learning. The model is trained on correct code examples and generates reliable translations, maintaining operational correctness while adapting to users with varying expertise levels who cannot write complex code themselves.
Solution Approach 2:
The patent implements feedback mechanisms where user interactions and corrections are used to continuously train and improve the machine learning model. This feedback loop enhances the system's ability to generate correct code for diverse user needs and languages, improving both reliability and adaptability over time as the model learns from real-world usage patterns.
3Measurement precision
If users must have deep knowledge of programming language syntax to perform data operations, then precise control is possible, but the ability to store, process, retrieve or analyze digital data is greatly limited
Solution Approach 1:
The patent substitutes the manual process of writing precise code with an automated machine learning system that generates correct code from natural language. This allows users without programming knowledge to perform precise data operations by simply describing what they want to achieve in natural language, dramatically increasing productivity while maintaining control precision through the intelligent translation system.
Data Source
AI summary
Implementations of this disclosure provide a machine learning model training system that receives user input being a natural language description of a search query, and packages and transmits the natural language description as a prompt to a plurality of large learning models (LLMs). The model training system also receives response from the plurality of LLMs being translations of the natural language descriptions to an executable search query and displays the translations to a user via a graphical user interface. The model training system receives user feedback via the graphical user interface that corresponds to indications as to whether each translation is correct, syntactically and/or semantically, and, in some examples, an indication of which response was preferred. The model training system also generates training data from the user input, translations generated by the plurality of LLMs, and user feedback, and subsequently, initiates training of a LLM using the training data.