Structured Data Normalization for Generative AI Model Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Training generative AI models using unstructured information can be inefficient and result in inaccurate models, especially in domains where most information is structured.
Innovation Solution
A computing platform obtains structured historical information, trains a foundational AI model, selects features from this model, normalizes the corresponding historical information, and uses it to train specialized generative AI models, which can then generate responses to user prompts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If generative AI models are trained using unstructured information, then the training process is simple to implement, but the training efficiency is low and model accuracy is poor
Solution Approach 1:
The patent transforms the training data from unstructured to structured format, changing the fundamental parameter of data organization. This transformation enables more efficient processing and training while maintaining implementation feasibility through automated structuring processes
Solution Approach 2:
The patent introduces an intermediary structuring layer that converts unstructured information into structured formats before training. This intermediary process bridges the gap between simple implementation and high efficiency by organizing data into manageable structures that accelerate training without requiring complete manual intervention
2Ease of manufacture
If generative AI models are trained using unstructured information, then the implementation is straightforward, but the model accuracy is inaccurate
Solution Approach 1:
The patent changes the data structure parameter from unstructured to structured, which directly improves model accuracy by enabling the model to learn from organized, relational data while keeping the implementation process relatively simple through automated transformation tools
Solution Approach 2:
The patent performs preliminary structuring of the training data before the actual model training begins. This preliminary organization of information into structured formats ensures higher accuracy from the start of training while maintaining ease of implementation through pre-processing automation
3Manufacturing precision
If structured historical information is used to train foundational AI models with feature selection and normalization, then the training accuracy is improved, but the training complexity increases
Solution Approach 1:
The patent segments the training process into distinct stages: foundational model training with structured data, feature selection, data normalization, and domain-specific model training. This segmentation manages complexity by breaking down the complex training pipeline into manageable, sequential steps that can be independently optimized
Solution Approach 2:
The patent implements automated feature selection and normalization processes that self-adjust based on the data characteristics. These self-service mechanisms reduce manual intervention and complexity while maintaining high training accuracy through algorithmic optimization
Data Source
AI summary
A computing platform may obtain, from an information storage source, historical information, which may be structured rather than unstructured. The computing platform may train, using the historical information, a foundational artificial intelligence (AI) model. The computing platform may select one or more features of the foundational AI model for use in training a generative AI model. The computing platform may identify a portion of the historical information corresponding to the selected one or more features. The computing platform may normalize the portion of the historical information. The computing platform may train, using the normalized portion of the historical information, the generative AI model. The computing platform may receive, from a user device, a generative AI prompt. The computing platform may generate, by inputting the generative AI prompt into the generative AI model, a generative AI response. The computing platform may send, to the user device, the generative AI response.