DomiKnowS: Declarative Knowledge Integration with Deep Neural Models

DomiKnowS is a Python library that facilitates the integration of domain knowledge in deep learning architectures. With DomiKnowS, you can express the structure of your data symbolically via graph declarations and seamlessly add logical constraints over outputs or latent variables to your deep models. This allows you to define domain knowledge explicitly, improving your models' explainability, performance, and generalizability, especially in low-data regimes.

While several approaches for integrating symbolic and sub-symbolic models have been introduced, no generic library facilitates programming for such integration with various underlying algorithms. DomiKnowS aims to simplify the programming for knowledge integration in training and inference phases while separating the knowledge representation from learning algorithms.

Contents

• Getting Started: Provides detailed instructions on how to get started with DomiKnowS, including installation, setting up the environment, and basic usage.
• Example Tasks: Contains examples that demonstrate the usage of DomiKnowS for various tasks, such as image classification, sequence modeling, and reinforcement learning. ( For more example see Examples Branch )
• Documentation: Provides comprehensive documentation on the DomiKnowS, including classes, methods, and their usage.
• Contributing: Explains how you can contribute to the development of DomiKnowS, including reporting issues, suggesting enhancements, and submitting pull requests.
• License: Contains information about the license of DomiKnowS and its terms of use.
• DomiKnowS Website: Contains documentation, example links, and an introductory video to DomiKnowS

Quick Start

1. Install DomiKnowS using pip install DomiKnowS.
2. Install Gurobi following the instructions here.
3. Refer to the Getting Started documentation for detailed instructions on how to define graph declarations, model declarations, initialize programs, and compose and execute programs using DomiKnowS.

Publications

• DomiKnowS: A Library for Integration of Symbolic Domain Knowledge in Deep Learning
• GLUECons: A Generic Benchmark for Learning Under Constraints

Acknowledgements

DomiKnowS is developed and maintained by HLR. We would like to acknowledge the contributions of the open-source community and express our gratitude to the developers of Gurobi for their excellent optimization solver.

Citation

If you use DomiKnowS in your research or work, please cite our paper:

@inproceedings{rajaby-faghihi-etal-2021-domiknows,
    title = "{D}omi{K}now{S}: A Library for Integration of Symbolic Domain Knowledge in Deep Learning",
    author = "Rajaby Faghihi, Hossein  and
      Guo, Quan  and
      Uszok, Andrzej  and
      Nafar, Aliakbar  and
      Kordjamshidi, Parisa",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
    month = nov,
    year = "2021",
    address = "Online and Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.emnlp-demo.27",
    doi = "10.18653/v1/2021.emnlp-demo.27",
    pages = "231--241",
}