Automatic Class Diagram Generation from Turkish User Stories
Context and Motivation: Class diagrams are widely adopted by the software development industry to describe either the inner structure of an object-oriented software or to describe the concepts and the relations among them in a domain. Even though the learning curve of drawing class diagrams is quite flat, it still takes time and effort to build them. User stories are used in agile software development to capture the needs and wishes of the users. They are widely available and easy to parse since they follow a template.
Problem: User stories include information about a system-to-be, yet it is difficult to get the high-level picture of the system by just reading a list of user stories. A picture might worth a thousand words and automaticly creating diagrams from user stories saves time and effort and increases the productivity of a software development team.
Goal: The goal of this project is to automatically generate domain models that describe the needs and wishes of the users from Turkish user stories. The pipeline is roughly as follows:
- Parse and understand user stories: Who are the users, which actions are done, what are the concepts that are operated on?
- Connect the concepts and relations, that is, build the structure of a domain model
- Visualize the diagram by either using state-of-the-art javascript libraries or static visualization libraries such as graphviz.
Skills required for the project:
- Basic understanding of Turkish language
- Ability to learn Natural Language Processing tools for Turkish (such as Zemberek, or pre-trained BERT models)
- Being able to work as a team and meet regularly with the supervisor