摘要
[目的/意义]实体语义关系分类是信息抽取重要任务之一,将非结构化文本转化成结构化知识,是构建领域本体、知识图谱、开发问答系统、信息检索系统的基础工作。[方法/过程]本文详细梳理了实体语义关系分类的发展历程,从技术方法、应用领域两方面回顾和总结了近5年国内外的最新研究成果,并指出了研究的不足及未来的研究方向。[结果/结论]热门的深度学习方法抛弃了传统浅层机器学习方法繁琐的特征工程,自动学习文本特征,实验发现,在神经网络模型中融入词法、句法特征、引入注意力机制能有效提升关系分类性能。
[Purpose/Significance]Entity semantic relation classification is one of the important tasks of information extraction,translate unstructured text into structured knowledge,the basic work of constructing domain ontology,knowledge graph,developing question answering system and information retrieval.[Method/Process]This paper sorted the development of entity semantic relation classification in detail,reviewed and summarized the latest research results in recent five years from technical method and application,finally pointed out the shortcomings of the research on entity semantic relation classification and the future research direction.[Result/Conclusion]Deep learning abandoned traditional machine learning methods with cumbersome feature engineering,automatic learning text features,the experiments showed that incorporating lexical and syntactic features into the neural network and introducing attention mechanism could effectively improve the performance of relation classification.
作者
李枫林
柯佳
Li Fenglin;Ke Jia(Department of Information Management,Wuhan University,Wuhan 430072,China)
出处
《现代情报》
CSSCI
2019年第2期47-56,84,共11页
Journal of Modern Information
关键词
实体语义关系
关系分类
神经网络
深度学习
entity semantic relation
relation classification
neural network
deep learning
