GLM-Based Test Generation for Software Requirement Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Software testing is a labor-intensive process that requires extensive human effort, as engineers manually review documentation to create test cases for each software requirement, which is burdensome and time-consuming.
Innovation Solution
Automated software program test generation using generative language models (GLMs) that identify relevant program segments and create test specifications and implementations based on requirement information, reducing the need for manual effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated test generation using GLMs is implemented, then productivity and time efficiency are improved, but device complexity increases due to integration of AI models
Solution Approach 1:
The patent introduces an intermediary layer between the GLM and the test generation process. This intermediary processes the model's outputs and transforms them into structured test cases, effectively mediating between the complex AI model and the simple test generation task. This allows the system to leverage the productivity benefits of GLMs while managing complexity through a dedicated translation layer.
Solution Approach 2:
The test generation process is segmented into distinct modules: requirement analysis, test case generation, test data creation, and execution. Each module can be independently optimized and maintained. The GLM is applied specifically to the generation phase rather than the entire testing workflow, reducing overall system complexity while maintaining productivity gains in critical paths.
2Measurement precision
If manual test case creation is performed, then measurement precision and test quality are improved, but loss of time increases due to extensive manual effort
Solution Approach 1:
The system performs preliminary analysis of requirements and program documentation using the GLM before actual test execution. Test cases are pre-generated and validated against requirements in advance, allowing manual reviewers to focus only on critical verification steps rather than creating tests from scratch. This preliminary automation maintains precision while dramatically reducing time investment.
Solution Approach 2:
The system implements feedback loops where test results are automatically analyzed and used to refine future test case generation. The GLM learns from previous testing outcomes and improves its precision over time. This continuous feedback mechanism allows the automated system to maintain or improve test accuracy while reducing the time required for each testing cycle.
Data Source
AI summary
Embodiments of the present disclosure relate to applications, platforms, architecture, etc. for automating software requirement verification. In particular, one or more generative language model (GLM) prompts may be generated based at least on program information that describes a software program and based on requirement information that corresponds to a requirement of the software program. Based on such prompts, the GLM may be able to automatically identify segments of the software program information that relate to the requirement. Further, based on the identified segments and the GLM prompts, the GLM may be able to automatically create (e.g., based on one or more additional prompts) testing architecture that may be used to verify whether the software program satisfies the requirement.