阿斯顿
commited on
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
- zh
|
| 5 |
+
tags:
|
| 6 |
+
- uml
|
| 7 |
+
- software-engineering
|
| 8 |
+
- requirements-modeling
|
| 9 |
+
- glm4
|
| 10 |
+
- code-generation
|
| 11 |
+
license: apache-2.0
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# AUG - Automated UML Model Generation
|
| 15 |
+
|
| 16 |
+
AUG (Automated UML Model Generation) is a fine-tuned version of GLM4-9B specifically designed for generating UML diagrams from natural language requirements. It supports class diagrams, use case diagrams, and sequence diagrams generation with high accuracy.
|
| 17 |
+
|
| 18 |
+
## Model Description
|
| 19 |
+
|
| 20 |
+
- **Model Type:** Fine-tuned GLM4-9B
|
| 21 |
+
- **Language(s):** English
|
| 22 |
+
- **Task:** UML Diagram Generation
|
| 23 |
+
- **License:** Apache 2.0
|
| 24 |
+
|
| 25 |
+
### Training Data
|
| 26 |
+
|
| 27 |
+
The model was trained on a comprehensive dataset of software requirements and their corresponding UML diagrams, focusing on:
|
| 28 |
+
|
| 29 |
+
- Class diagrams
|
| 30 |
+
- Use case diagrams
|
| 31 |
+
- Sequence diagrams
|
| 32 |
+
|
| 33 |
+
## Performance and Limitations
|
| 34 |
+
|
| 35 |
+
### Metrics
|
| 36 |
+
|
| 37 |
+
The model achieves the following performance metrics:
|
| 38 |
+
|
| 39 |
+
- Precision: 78.68%
|
| 40 |
+
- Recall: 67.37%
|
| 41 |
+
- F1 Score: 72.59%
|
| 42 |
+
|
| 43 |
+
Compared to ChatGPT 4.0, AUG shows improved performance in:
|
| 44 |
+
|
| 45 |
+
- Recall rate
|
| 46 |
+
- Overall F1 score
|
| 47 |
+
|
| 48 |
+
### Limitations
|
| 49 |
+
|
| 50 |
+
While the model performs well, users should be aware that:
|
| 51 |
+
|
| 52 |
+
- The model's output may still require human verification
|
| 53 |
+
- Complex software architectures might need additional refinement
|
| 54 |
+
- Performance may vary based on the clarity of input requirements
|
| 55 |
+
|
| 56 |
+
## Training Procedure
|
| 57 |
+
|
| 58 |
+
The model was fine-tuned from GLM4-9B using:
|
| 59 |
+
|
| 60 |
+
- Specialized UML diagram dataset
|
| 61 |
+
- Requirements-to-diagram mapping
|
| 62 |
+
- Clarification dialogue patterns
|
| 63 |
+
|
| 64 |
+
## Additional Information
|
| 65 |
+
|
| 66 |
+
- **GitHub Repository:** https://github.com/XIAOLingQ/AUG
|
| 67 |
+
- **Demo Video:** https://youtu.be/kHbCPK6kOag
|
| 68 |
+
|
| 69 |
+
## Acknowledgements
|
| 70 |
+
|
| 71 |
+
This model was developed by researchers from Wuhan Textile University and Wuhan University. Special thanks to all contributors and the open-source community.
|