Automated Training Data Generation Using Template-Based Copying
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The inefficiency and cost associated with manually creating high-quality guide text and result text as training sets for language models, which affects the quality of the final output content.
Innovation Solution
A method and apparatus for generating training data that involves acquiring sample data, determining a data generation template, generating question data and answer data, and combining them into training data, thereby automating the process and reducing manual effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If training data is written manually, then high quality guide text and result text can be obtained, but the process is inefficient and costly
Solution Approach 1:
The patent uses template-based copying to generate training data. Pre-defined templates with placeholders are copied and filled with actual data from sample records, automatically generating guide text and result text pairs without manual writing, thus improving efficiency while maintaining quality through structured template design
Solution Approach 2:
The system performs self-service by automatically generating training data from existing sample data using templates. The generation model and template engine work autonomously to create guide text and result text pairs without requiring manual intervention for each training sample, reducing both time and cost
2Manufacturing precision
If training data is written manually, then high quality guide text and result text can be obtained, but costs increase
Solution Approach 1:
By copying and filling templates with sample data, the system eliminates expensive manual writing processes. The template-based approach reuses proven structures multiple times, reducing the labor cost associated with creating each training sample while preserving quality through consistent template design
Solution Approach 2:
The system changes the parameter of data generation from manual text creation to automated template filling. This parameter change transforms a high-cost process into a low-cost automated process, maintaining output quality through structured templates while dramatically reducing the energy and resource expenditure required
3Manufacturing precision
If manual writing is used for training sets, then quality control is possible, but the process is time-consuming
Solution Approach 1:
Templates are prepared in advance with predefined structures, placeholders, and formatting rules. This preliminary action ensures that when training data is generated, the quality standards are already embedded in the templates, eliminating the need for time-consuming manual quality control while maintaining high quality output
Solution Approach 2:
By copying proven template structures and filling them with sample data, the system generates multiple training samples simultaneously through automated processing. This copying approach maintains quality consistency across all generated samples while reducing the time required compared to manual writing of each sample individually
Data Source
AI summary
A method of generating training data, a readable medium and an electronic device are provided. The method includes: acquiring sample data; determining a data generation template according to the sample data; generating question data according to the first data in the sample data and the data generation template, and determining answer data according to the second data other than the first data in the sample data; and combining the question data and the answer data into the training data.