Tool4Boxology is a modular visual toolkit for designing and validating hybrid AI systems using the Boxology methodology. It provides a shared conceptual and visual language to describe systems that combine symbolic reasoning and machine learning.
This project is inspired by the work of Frank van Harmelen et al. in their paper:
"Modular Design Patterns for Hybrid Learning and Reasoning Systems: A Taxonomy, Patterns and Use Cases" (Web Semantics, 2023)
Hybrid AI systems lack a co 8000 mmon language for design, validation, and communication. Tool4Boxology addresses this by:
- Providing a visual grammar and formal syntax for system components
- Enabling modular, explainable design with reusable patterns
- Offering real-time validation tools to reduce design errors early
- Enhancing documentation, automation, and traceability
- โ Draw.io Plugin with validation logic
- ๐ง Formal grammar for elementary Boxology patterns
- ๐งฐ Custom vocabulary libraries
- ๐ฅ๏ธ GoJS-based interface for direct interaction
- ๐ณ Dockerized setup for reproducible deployment
| Folder | Description |
|---|---|
Boxology-Docker |
Docker container setup based on fjudith/drawio, including preloaded plugins and custom libraries. |
Boxology-interface |
Custom visual interface for Boxology models using GoJS. |
Boxology-plugin |
Draw.io plugin and vocabulary library for manual integration. |
ElementaryPattern |
Elementary patterns written in DOT language for modular visualization. |
Report |
Development report tracing project progress from scratch to implementation. |
- Open Draw.io (web or desktop)
- Import the plugin from the
Boxology-pluginfolder - Load the vocabulary
.drawiofile to access custom shapes
git clone https://github.com/SDM-TIB/Tool4Boxology.git
cd Tool4Boxology/Boxology-Docker
docker-compose upAccess Draw.io at http://localhost:8080 with Boxology support.
Open the GoJS-based visual editor from the Boxology-interface folder. Instructions are included in its README.
A sample hybrid AI pipeline using Boxology grammar. This diagram is a use case from Boxology paper (Robot in Action).
โ Check for validation!
- RDF export for integration into knowledge graphs
- SHACL-based validation for semantic consistency
- Enhanced support for reasoning templates and ontology-based rules
- Harmelen, F. van, Liao, B., Ifrim, G., & Groth, P. (2023). Modular Design Patterns for Hybrid Learning and Reasoning Systems. Journal of Web Semantics, 79, 100769. DOI
- Keet, C. M., & Rodrรญguez-Muro, M. (2023). Combining Machine Learning and Semantic Web: A Systematic Mapping Study. ACM Computing Surveys, 55(7), Article 140.
๐ง Mahsa Forghani Tehrani
๐ Master Student, Leibniz University Hannover
๐ฎ mahsa.forghani.tehrani@stud.uni-hannover.de
- ๐ง Code is licensed under the MIT License.
- ๐ Documentation, diagrams, and examples are licensed under the Creative Commons Attribution 4.0 International License.
Pull requests and feedback are welcome! Please open issues for bugs, ideas, or questions.