Abstract

Knowledge graph embedding can learn low-rank vector representations for knowledge graph entities and relations, and has been a main research topic for knowledge graph completion. Several recent works suggest that convolutional neural network (CNN)-based models can capture interactions between head and relation embeddings, and hence perform well on knowledge graph completion. However, previous convolutional network models have ignored the different contributions of different interaction features to the experimental results. In this paper, we propose a novel embedding model named DyConvNE for knowledge base completion. Our model DyConvNE uses a dynamic convolution kernel because the dynamic convolutional kernel can assign weights of varying importance to interaction features. We also propose a new method of negative sampling, which mines hard negative samples as additional negative samples for training. We have performed experiments on the data sets WN18RR and FB15k-237, and the results show that our method is better than several other benchmark algorithms for knowledge graph completion. In addition, we used a new test method when predicting the Hits@1 values of WN18RR and FB15k-237, named specific-relationship testing. This method gives about a 2% relative improvement over models that do not use this method in terms of Hits@1.

Highlights

  • Knowledge graphs are usually expressed in a highly structured form, where nodes denote the entities and edges represent different relations entities

  • Based on the above observations, we propose a dynamic convolution network-based model using a method of mining hard negative samples for knowledge graph completion; we name it DyConvNE

  • We propose a new model, called DyConvNE, based on a dynamic convolution network, which uses dynamic convolution to dynamically assign weights to the interaction features of the extracted entities and relationship embeddings; We propose a method to mine hard negative samples and demonstrate the effectiveness of the method through ablation experiments; We use specific-relationship-testing to obtain better performance on Hits@1; We conduct some experiments to evaluate the performance of the proposed method

Read more

Summary

Introduction

Knowledge graphs are usually expressed in a highly structured form, where nodes denote the entities and edges represent different relations entities. In the field of knowledge graph completion, traditional negative sampling is usually used to randomly replace the head entity or tail entity of the positive triples to obtain a certain number of negative triples before each round of model training, and these positive triples and negative triples are put together to train the model. We propose a new model, called DyConvNE, based on a dynamic convolution network, which uses dynamic convolution to dynamically assign weights to the interaction features of the extracted entities and relationship embeddings; We propose a method to mine hard negative samples and demonstrate the effectiveness of the method through ablation experiments; We use specific-relationship-testing to obtain better performance on Hits@1; We conduct some experiments to evaluate the performance of the proposed method. Experimental results demonstrate that our method obtains competitive performance on both WN18RR and FB15k-237

Related Work
Our Approach
Definition
Our Model
Dynamic Convolution
Mining Hard Negative Samples
Training Objective
Experiments
Datasets
Experimental Setup
Main Results
Specific Relationship Testing
Case Study
Ablation Study
Hard Negative Sampling Study
Conclusions

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.