Completing Taxonomies with Relation-Aware Mutual Attentions
Completing Taxonomies with Relation-Aware Mutual Attentions
复制标题
DOI：
--
复制
发表时间：
2023
期刊：
影响因子：
--
通讯作者：
Qingkai Zeng;Zhihan Zhang;Jinfeng Lin;Meng Jiang
中科院分区：
文献类型：
--
作者：
Qingkai Zeng;Zhihan Zhang;Jinfeng Lin;Meng Jiang
分类法通过知识的结构表示为许多应用提供服务。为了将新兴概念纳入现有分类法，分类法完成的任务旨在为新兴查询概念找到合适的位置。之前的工作使用预训练的语言模型捕获了查询概念术语和定义串联内的同质标记级交互。然而，他们忽略了查询概念及其相关概念的术语和定义之间的标记级交互。在这项工作中，我们建议捕获具有不同类型关系的概念的不同文本组件之间的异构标记级交互。我们设计了一个关系感知相互关注模块（RAMA）来学习这种交互以完成分类。实验结果表明，我们基于 RAMA 的新分类完成框架在六个分类数据集上实现了最先进的性能。本文属于“应用与分析——知识图谱构建”，属于“新颖研究论文”类别。
Taxonomies serve many applications with a structural representation of knowledge. To incorporate emerging concepts into existing taxonomies, the task of taxonomy completion aims to find suitable positions for emerging query concepts. Previous work captured homogeneous token-level interactions inside a concatenation of the query concept term and definition using pre-trained language models. However, they ignored the token-level interactions between the term and definition of the query concepts and their related concepts. In this work, we propose to capture heterogeneous token-level interactions between the different textual components of concepts that have different types of relations. We design a relation-aware mutual attention module (RAMA) to learn such interactions for taxonomy completion. Experimental results demonstrate that our new taxonomy completion framework based on RAMA achieves the state-of-the-art performance on six taxonomy datasets. This paper belongs to “Application and analysis - Knowledge Graph Construction”, and in the “ Novel research paper” category.