Automated UML class diagram generation from textual requirements using NLP techniques
Translating textual requirements into precise Unified Modeling Language (UML) class diagrams poses challenges due to the unstructured and often ambiguous nature of text, which can lead to inconsistencies and misunderstandings during the initial stages of software development. Current methods often s...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Politeknik Negeri Padang
2024
|
| Online Access: | http://psasir.upm.edu.my/id/eprint/114504/1/114504.pdf http://psasir.upm.edu.my/id/eprint/114504/ https://joiv.org/index.php/joiv/article/view/3482 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Translating textual requirements into precise Unified Modeling Language (UML) class diagrams poses challenges due to the unstructured and often ambiguous nature of text, which can lead to inconsistencies and misunderstandings during the initial stages of software development. Current methods often struggle with effectively addressing these challenges due to limitations in handling diverse and complex textual requirements, which may result in incomplete or inaccurate UML diagrams. This study aims to propose a Natural Language Processing (NLP) model that analyzes and comprehends textual requirements to extract relevant information for generating UML class diagrams, ensuring accuracy and consistency between the diagrams and requirement descriptions. The research employs a four-step approach: preprocessing to handle text noise and redundancy, sentence classification to distinguish between "class" and "relationship" sentences, syntactic analysis to examine grammatical structures, and UML class diagram generation based on predefined rules. The results show that the model achieved a classification accuracy of 88.46% with a high Area Under the Curve (AUC) value of 0.9287, indicating robust performance in distinguishing between class definitions and relationships. This study highlights that existing methods may not fully address the nuances of translating complex textual requirements into accurate UML diagrams. This study successfully demonstrates an automated method for generating UML class diagrams from textual requirements and suggests that future research could expand datasets, optimize feature extraction, explore advanced models, and develop automated rule generation methods for further improvements. |
|---|