About & contributing

Why a unified EA repo?

Entity-alignment research moves fast, but its code does not travel well: every method ships its own data reader, training loop and evaluation script, often in incompatible frameworks. That makes fair comparison and onboarding unnecessarily painful.

EntityAlignment-Nexus exists to serve the entity-alignment community by collecting models behind a single clean engine, so that:

• a newcomer can read one method and understand the rest;
• a researcher can reproduce a number with one command;
• a practitioner can compare methods on equal footing.

Design principles

• One engine, many models. A shared data.py, trainer.py and metrics.py; each model is just an encoder + loss + a YAML config.
• Read it or run it. Every model exists both as a package module and as a self-contained, documented notebook.
• Honest reporting. Where we fall short of a paper, the gap is stated and explained rather than hidden.
• Reproducible by construction. Every run snapshots its exact config and writes logs, CSVs, curves and checkpoints.

How to add a model

%%{init: {'theme':'base','themeVariables':{'fontSize':'14px','fontFamily':'Inter, sans-serif','lineColor':'#7d8590','primaryTextColor':'#e6edf3'}}}%%
flowchart LR
    S(["pick a model"]):::start
    subgraph CORE["1 . implement"]
        direction TB
        A["models/your_model.py<br/><i>encoder + loss</i>"]:::model
        B["trainer.py<br/><i>reuse or add a trainer</i>"]:::train
        C["configs/your_model.yaml<br/><i>hyper-parameters</i>"]:::cfg
        A --> B --> C
    end
    subgraph SHARE["2 . document"]
        direction TB
        D["Notebook/your_model.ipynb<br/><i>self-contained, readable</i>"]:::nb
        E["docs/models/your_model.md<br/><i>+ register in mkdocs.yml</i>"]:::doc
        D --> E
    end
    S --> CORE --> SHARE --> PR(["open a pull request"]):::done
    style CORE fill:#0d1117,stroke:#30363d,color:#e6edf3
    style SHARE fill:#0d1117,stroke:#30363d,color:#e6edf3
    classDef start fill:#1f2937,stroke:#8b949e,stroke-width:2px,color:#e6edf3;
    classDef model fill:#0c2d6b,stroke:#58a6ff,stroke-width:2px,color:#dbeafe;
    classDef train fill:#3b0764,stroke:#a371f7,stroke-width:2px,color:#f3e8ff;
    classDef cfg   fill:#3a2a05,stroke:#d29922,stroke-width:2px,color:#fde68a;
    classDef nb    fill:#7c2d54,stroke:#f778ba,stroke-width:2px,color:#ffe4f0;
    classDef doc   fill:#14532d,stroke:#3fb950,stroke-width:2px,color:#dcfce7;
    classDef done  fill:#238636,stroke:#3fb950,stroke-width:2px,color:#ffffff;

1. Model - add code/src/models/<your_model>.py with the encoder and its loss functions.
2. Trainer - reuse an existing trainer or add one in code/src/trainer.py.
3. Config - add configs/<your_model>.yaml.
4. Notebook - mirror an existing notebook so the method can be read inline.
5. Docs - add docs/models/<your_model>.md following the page template, and register it in mkdocs.yml.

Building the docs

pip install -r requirements-docs.txt
mkdocs serve          # local preview
mkdocs build          # static site into site/

The site deploys to GitHub Pages automatically on push to main (.github/workflows/docs.yml).

Citation

@software{ea_dbp15k,
  title  = {EntityAlignment-Nexus: A Unified Framework of Entity Alignment Models on DBP15K},
  author = {Nadjib ZAHAF},
  year   = {2026},
  url    = {https://github.com/Z-Nadjib/EntityAlignment-Nexus}
}

Please also cite the original paper of any model you use - links are on each model page.

License

Released under the MIT License. The original papers and the DBP15K benchmark remain the property of their respective authors.

Acknowledgements

Thanks to the authors of NAEA, BootEA, AliNet, KECG, GCN-Align, JAPE, DGMC, MRAEA and RREA, and to the maintainers of the DBP15K benchmark and the OpenEA study.