• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>IEEE International Conference on Artificial Intelligence and Industrial Design >A Zero-shot Entity Knowledge Representation Learning Method based on Self-attention Mechanism
【24h】

A Zero-shot Entity Knowledge Representation Learning Method based on Self-attention Mechanism

机译：基于自我关注机制的零射实体知识表示学习方法

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Zero-shot entity knowledge representation learning method can add new entities to the knowledge graph while avoiding training all data, which has been one of research hotspots in the field of knowledge graph. In view of the problem that the existing method only averages the knowledge representations of related entities as output and ignores the mutual influence between them, a zero-shot entity knowledge representation learning method based on self-attention mechanism, namely, ZKRL, is proposed. The method adopts a self-attention mechanism to learn dependencies between related entities and an attention mechanism to estimate contribution of different related entities, successfully capturing internal structural characteristics and effectively improving expressive ability of knowledge representation. Through link prediction and triplet classification experiments on zero-shot datasets WN1000 and WN3000, the remarkable improvement of ZKRL is validated.
机译：零拍实体知识表示学习方法可以向知识图中添加新实体，同时避免培训所有数据，这是知识图类领域的研究热点之一。 鉴于现有方法仅平均相关实体的知识表示作为输出并忽略它们之间的相互影响，提出了一种基于自我关注机制，即ZKRL的零射实体知识表示学习方法。 该方法采用自我关注机制来学习相关实体与注意机制之间的依赖性，以估计不同相关实体的贡献，成功地捕获内部结构特征，有效地提高知识表示的表达能力。 通过对零拍数据集WN1000和WN3000的链路预测和三联分类实验，验证了ZKRL的显着改进。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号