DomiKnowS-NLP

Video Link

  • CONLL Example In the Video: Run the example shown in the video in google colab.
  • Getting Started: Provides detailed instructions on how to get started with DomiKnowS, including installation, setting up the environment, and basic usage.
  • Documentation: Provides comprehensive documentation on the DomiKnowS, including classes, methods, and their usage.
  • Walkthough Example: 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 )
  • Tutorial Examples: Simple and diverse examples are outlined with detailed explanations in a Jupyter file to run.
  • FAQ: Read our FAQ file if you have any questions.
  • License: Contains information about the license of DomiKnowS and its terms of use.
  • Report Issues: Encounter a problem? Let us know by submitting an issue.
  • Suggest Enhancements: Have an idea to make DomiKnowS better? Share your suggestion.
  • Submit Pull Requests: Ready to contribute code or features directly? Create a pull request.

Abstract

This library is for designing deep learning architectures for the problems that have structured output when this structure can be expressed symbolically using logical constraints between output concepts. Our showcases mostly focus on NLP problems. Though the language models have taken over in many NLP tasks, using explicit knowledge about the domain or world can be helpful for generalizability beyond the observed data and explain-ability of the decisions. Though there are several approaches for such an integration of symbolic and subsymbolic models, currently, there is no library to facilitate the programming for such an integration in a generic way and the models need to be hard coded task-specifically. Our library aims to facilitate programming for such an integration.