References & Citations
Computer Science > Computation and Language
Title: DisenE: Disentangling Knowledge Graph Embeddings
(Submitted on 28 Oct 2020 (v1), last revised 12 Nov 2020 (this version, v2))
Abstract: Knowledge graph embedding (KGE), aiming to embed entities and relations into low-dimensional vectors, has attracted wide attention recently. However, the existing research is mainly based on the black-box neural models, which makes it difficult to interpret the learned representation. In this paper, we introduce DisenE, an end-to-end framework to learn disentangled knowledge graph embeddings. Specially, we introduce an attention-based mechanism that enables the model to explicitly focus on relevant components of entity embeddings according to a given relation. Furthermore, we introduce two novel regularizers to encourage each component of the entity representation to independently reflect an isolated semantic aspect. Experimental results demonstrate that our proposed DisenE investigates a perspective to address the interpretability of KGE and is proved to be an effective way to improve the performance of link prediction tasks.
Submission history
From: Xiaoyu Kou [view email][v1] Wed, 28 Oct 2020 03:45:19 GMT (404kb,D)
[v2] Thu, 12 Nov 2020 12:53:03 GMT (0kb,I)
Link back to: arXiv, form interface, contact.