| 陈曦,陈华钧,张文.规则增强的知识图谱表示学习方法[J].情报工程,2017,3(1):026-034 |
| 规则增强的知识图谱表示学习方法 |
| Rule-Augmented Representation Learning Approach in Knowledge Graph |
| |
| DOI：10.3772/j.issn.2095-915x.2017.01.003 |
| 中文关键词: 知识图谱，表示学习，规则挖掘，推理 |
| 英文关键词: Knowledge graph, representation learning, rule mining, reasoning |
| 基金项目: |
| 作者 | 单位 | | 陈曦 | 浙江大学计算机科学与技术系 | | 陈华钧 | 浙江大学计算机科学与技术系 | | 张文 | 浙江大学计算机科学与技术系 |
|
| 摘要点击次数: 6473 |
| 全文下载次数: 5716 |
| 中文摘要: |
| 知识图谱（Knowledge Graph，简称KG）的表示学习方法旨在将知识图谱的实体和关系表示为稠密低维实值向量, 进而在低维向量空间中高效计算实体、关系及其之间的复杂语义关联, 在知识图谱的构建、推理、融合、挖掘以及应用中具有重要作用。已有的知识图谱表示方法仅仅考虑了知识图谱中的直接事实，忽略了知识图谱中一些隐藏的语义信息，这些语义信息对于知识图谱关系和实体的嵌入表示有着重要的影响。本文提出了一种规则增强的知识图谱表示学习方法，该方法首先通过知识图谱规则挖掘的方法提取一组可代表知识图谱语义信息的Horn 逻辑规则，随后通过基于规则的物化推理方法将相应的隐藏语义信息注入到知识图谱表示学习模型中。实验结果表明，基于规则增强的方法可以显著提升已有知识图谱表示学习模型在链接预测和定理预测上的效果和性能。 |
| 英文摘要: |
| Representation learning in Knowledge Graph (KG) aims to project the entities and relations into a dense, real-valued and low-dimensional vectors, so as to efficiently measure complex semantic correlations between entities and relations, and plays a important role in knowledge acquisition, inference, fusion, mining and applications of KG. Existing methods of representation learning in KG only concern direct facts in KG,ignoring some implicit semantic information. The paper proposes proposed a rule-augmented representation learning approach in KG. This method firstly implemented a rule mining algorithm to extract Horn rules from the KB. Then, a rule-based materialization reasoning was used to encode logical rules into our learning models. Experimental results on real-world datasets showed that, this approach achieved significant and consistent improvements compared with these baselines in the link and axiom prediction tasks. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
