This repo contains the code for our ISWC paper "FeaBI: A Feature Selection-Based Framework for Interpreting KG Embeddings" by Youmna Ismaeil et al. .
Several methods for explaining predictions made by KG embeddings have been proposed in the literature. However, none of them targeted the problem of constructing model explanations for embeddings, i.e., interpretable KG representations that behave similarly to embeddings on certain tasks. We address this problem and propose a novel method for generating interpretable vectors for entity embeddings. To achieve this, we employ embedded feature selection techniques to ex-tract from the KG, on which the embedding model was trained, propositional features that are important for a given KG embedding model. Our approach sheds light on the information in the KG captured by embeddings and provides valuable insights that can be used to further enhance the embedding models. Additionally, we demonstrate the usefulness of our method for explaining embedding-based entity similarity.
This software is a research prototype, solely developed for and published as part of the above publication.
Thank you for your interest in our project! We are excited to announce that our code is currently under internal review. We understand that many of you are eager to access the code and we appreciate your patience. Rest assured that we are committed to delivering well-documented solution.
Once the code review is completed and the code is ready for public release, we will make it available on this GitHub repository. In the meantime, we encourage you to use the watch option for this repository to receive instant updates when the code becomes available.
FEABI is open-sourced under the AGPL-3.0 license. See the LICENSE file for details.