2018/08/13 by Dai Quoc Nguyen, Thanh Vu, Nguyen, Dai Quoc +7 · 2 citations
Computer Science · #Advanced Graph Neural Networks #Computation and Language (cs.CL) #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Recommender Systems and Techniques #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1808.04122
openalex publication_date 2018/08/13 · openalex created_date 2022/08/04 · openalex updated_date 2026/07/28
In this paper, we introduce an embedding model, named CapsE, exploring a\ncapsule network to model relationship triples (subject, relation, object). Our\nCapsE represents each triple as a 3-column matrix where each column vector\nrepresents the embedding of an element in the triple. This 3-column matrix is\nthen fed to a convolution layer where multiple filters are operated to generate\ndifferent feature maps. These feature maps are reconstructed into corresponding\ncapsules which are then routed to another capsule to produce a continuous\nvector. The length of this vector is used to measure the plausibility score of\nthe triple. Our proposed CapsE obtains better performance than previous\nstate-of-the-art embedding models for knowledge graph completion on two\nbenchmark datasets WN18RR and FB15k-237, and outperforms strong search\npersonalization baselines on SEARCH17.\n