DomiKnowS: A Library for Integration of Symbolic Domain Knowledge in Deep Learning

TL;DR

DomiKnowS is a library that enables seamless integration of symbolic domain knowledge into deep learning models, enhancing explainability, performance, and generalization, especially in low-data scenarios.

Contribution

The paper introduces DomiKnowS, a novel library that simplifies programming for integrating symbolic domain knowledge with deep learning models in a generic and flexible manner.

Findings

Improved model explainability and performance with symbolic knowledge integration

Enhanced generalization in low-data regimes

Validated on various NLP benchmark tasks

Abstract

We demonstrate a library for the integration of domain knowledge in deep learning architectures. Using this library, the structure of the data is expressed symbolically via graph declarations and the logical constraints over outputs or latent variables can be seamlessly added to the deep models. The domain knowledge can be defined explicitly, which improves the models' explainability in addition to the performance and generalizability in the low-data regime. Several approaches for such an integration of symbolic and sub-symbolic models have been introduced; however, there is no library to facilitate the programming for such an integration in a generic way while various underlying algorithms can be used. Our library aims to simplify programming for such an integration in both training and inference phases while separating the knowledge representation from learning algorithms. We showcase various NLP benchmark tasks and beyond. The framework is publicly available at Github(https://github.com/HLR/DomiKnowS).