Knowledge Graph (KG) Embedding methods have been widely studied in the past few years and many methods have been proposed. These methods represent entities and relations in the KG as vectors in a vector space, trained to distinguish correct edges from the incorrect ones. For this distinction, simple functions of vectors’ dimensions, called interactions, are used. These interactions are used to calculate the candidate tail entity vector which is matched against all entities in the KG. However, for most of the existing methods, these interactions are fixed and manually specified. In this work, we propose an automated framework for discovering the interactions while training the KG Embeddings. The proposed method learns relevant interactions along with other parameters during training, allowing it to adapt to different datasets. Many of the existing methods can be seen as special cases of the proposed framework. We demonstrate the effectiveness of the proposed method on link prediction task by extensive experiments on multiple benchmark datasets.
We study the problem of inducing interpretability in Knowledge Graph (KG) embeddings. Learning KG embeddings has been an active area of research in the past few years, resulting in many different models. However, most of these methods do not address the interpretability (semantics) of individual dimensions of the learned embeddings. In this work, we study this problem and propose a method for inducing interpretability in KG embeddings using entity co-occurrence statistics. The proposed method significantly improves the interpretability, while maintaining comparable performance in other KG tasks.
Knowledge Graph (KG) embedding has emerged as a very active area of research over the last few years, resulting in the development of several embedding methods. These KG embedding methods represent KG entities and relations as vectors in a high-dimensional space. Despite this popularity and effectiveness of KG embeddings in various tasks (e.g., link prediction), geometric understanding of such embeddings (i.e., arrangement of entity and relation vectors in vector space) is unexplored – we fill this gap in the paper. We initiate a study to analyze the geometry of KG embeddings and correlate it with task performance and other hyperparameters. To the best of our knowledge, this is the first study of its kind. Through extensive experiments on real-world datasets, we discover several insights. For example, we find that there are sharp differences between the geometry of embeddings learnt by different classes of KG embeddings methods. We hope that this initial study will inspire other follow-up research on this important but unexplored problem.