场景化知识图谱及构建方法
Scenario-Based Knowledge Graph and Construction Method
陆泉 1陈静宇 2陈帅朴 1姚苏梅 2陈静3
作者信息
- 1. 武汉大学信息资源研究中心,湖北武汉 430072;武汉大学大数据研究院,湖北武汉 430072
- 2. 武汉大学信息资源研究中心,湖北武汉 430072
- 3. 华中师范大学信息管理学院,湖北武汉 430079
- 折叠
摘要
[目的/意义]基于场景理论及领域知识图谱,提出场景化知识图谱的概念及内涵,描述知识的产生与应用场景信息,从而提高知识图谱应用的精准性和适用性.[方法/过程]首先,基于场景理论、知识场景化及知识图谱构建技术,提出了场景化知识图谱的定义及结构;其次,以医学场景化知识图谱为例,提出了场景化知识图谱的构建方法.具体而言,结合BiLSTM-CRF模型、UIEE模型、基于规则的方法构建了知识抽取及融合方法,借助Neo4j图数据库和Cypher查询语言实现场景化知识图谱的存储与查询.[结果/结论]首先,提出了场景化知识图谱是描述知识场景属性的知识图谱,总结出场景具有可融合性、可继承性和可推理性的特点;其次,以轻度认知障碍为例进行实证探究,结果表明当前不同类型的场景抽取性能参差不齐;各种类型的场景属性总体分布和在不同实体中的分布均不同;在所构建的场景化知识图谱中,场景属性的可融合性、可继承性和可推理性特征明显;在场景匹配的基础上进行查询能减少冗余和错误信息、提升结果的精准性和适用性.[创新/局限]本研究进一步丰富了领域知识图谱研究,通过描述知识的产生和应用场景提升知识图谱应用的精准性与适用性.
Abstract
[Purpose/significance]Based on scenario theory and domain knowledge graph,the concept and connotation of scenario-based knowledge graph are proposed to describe the generation and application scenario information of knowledge,thereby improving the precision and applicability of knowledge graph application.[Method/process]Firstly,based on scenario theory,knowledge scener-ization,and knowledge graph construction techniques,the definition and structure of scenario-based knowledge graph are proposed;Secondly,taking the medical scenario-based knowledge graph as an example,a method for constructing the scenario-based knowl-edge graph is proposed.Specifically,a knowledge extraction and fusion method is constructed by combining the BiLSTM-CRF model,the UIE model,and the rule-based method.The storage and query of scenario-based knowledge graph is implemented using the Neo4j graph database and the Cypher query language.[Result/conclusion]Firstly,it is proposed that the scenario-based knowledge graph is a knowledge graph that describes the attributes of knowledge scenes,and summarizes the characteristics of scene fusion,inheritance,and reasonability;Secondly,taking mild cognitive impairment as an example to conduct empirical research,the results show that the current performance of different types of scene extraction is uneven;The overall distribution of various types of scene attributes and their distribution in different entities are different;In the constructed scenario-based knowledge graph,the fusibility,inheritability,and reasonability of scene attributes are obvious;Searching based on scene matching can reduce redundancy and error information,and improve the precision and applicability of results.[Innovation/limitation]This study further enriches the research on domain knowledge graph,improving the precision and applicability of knowledge graph applications by describing the generation and applica-tion scenarios of knowledge.
关键词
场景化知识图谱/场景/知识图谱/知识融合/轻度认知障碍Key words
scenario-based knowledge graph/scenarios/knowledge graph/knowledge fusion/mild cognitive impairment引用本文复制引用
基金项目
国家社会科学基金重点项目(20ATQ008)
出版年
2024
NETL
NSTL
万方数据