Neural Transformer Unit Test Generation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Automated unit testing in software development is time-consuming and expensive, and existing methods struggle to generate correct and readable unit tests efficiently.

Innovation Solution

A unit test case generation system using a neural transformer model with attention, trained through transfer learning on both natural language text and source code, to predict unit test cases for methods in a programming language, capturing syntax and semantics to generate executable test statements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If automated unit testing is used to reduce manual testing effort, then testing productivity is improved, but the generated tests may lack correctness and readability

Engineering Contradiction:

Improvetesting productivityVSAvoidtest correctness

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent introduces a neural transformer model as an intermediary between the method code and the generated unit test. This intermediary translates the method into natural language descriptions and then generates corresponding JUnit test cases, improving both the correctness and readability of automated tests while maintaining high productivity

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces traditional mechanical testing approaches with an AI-based neural transformer system. Instead of relying on rule-based or template-based test generation, the system uses deep learning models trained on large datasets to generate high-quality unit tests automatically, achieving both efficiency and reliability

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Speed

If traditional automated testing tools are used, then testing speed is improved, but the quality and readability of generated tests deteriorate

Engineering Contradiction:

Improvetesting speedVSAvoidtest readability

Core Design Contradiction:

SpeedVSEase of operation

Solution Approach 1:

The neural transformer model serves as a mediator that first converts method code into natural language descriptions, then translates these descriptions into readable JUnit test cases. This two-stage translation process ensures both fast generation and high readability of the resulting tests

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system changes the parameter space by training the neural transformer on large datasets of method-code pairs and their corresponding natural language descriptions. This parameter optimization enables the model to generate tests that are both fast to produce and highly readable, resolving the contradiction between speed and quality

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12175214B2Unit test case generation with transformers

Publication Date: 2024.12.24 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

A unit test generation system employs a neural transformer model with attention to generate candidate unit test sequences given a focal method of a programming language. The neural transformer model is pre-trained with source code programs and natural language text and fine-tuned with mapped test case pairs. A mapped test case pair includes a focal method and a unit test case for the focal method. In this manner, the neural transformer model is trained to learn the semantics and statistical properties of a natural language, the syntax of a programming language and the relationships between the code elements of the programming language and the syntax of a unit test case.