基于频繁语义模式的主题标注--以肿瘤靶向治疗技术为例  

Topic Labeling Based on Frequent Semantic Patterns:A Case of the Tumor Targeted Therapy Technology

作　　者：于慧娴 沈漫竹 梁曦 赵迎迎 袁红梅[1] Yu Huixian

机构地区：[1]沈阳药科大学,辽宁沈阳110016

出　　处：《情报理论与实践》2022年第7期196-201,共6页Information Studies:Theory & Application

基　　金：沈阳药科大学工商管理学院学科建设课题“基于专利数据的制造产业技术情报分析”的研究成果,项目编号:2021-sygsxk-01。

摘　　要：[目的/意义]为提高主题建模结果的可读性,解决主题标注过程中主观性强、可解释性弱的问题,引入本体和关联规则构建频繁语义模式。[方法/过程]以肿瘤靶向治疗专利数据为研究对象,首先构建LDA2vec模型挖掘主题;然后考虑主题词的语义与共现,将UMLS本体与FP-growth关联规则相结合,得到频繁语义模式的主题标注;最后将结果与Canopy方法对比评估。[结果/结论]实证结果显示,基于频繁语义模式的标注能够从统计和语义的角度概括主题含义,在主题标注工作中效果更佳。[Purpose/significance]To improve the readability of topic modeling results,solve the problem of subjectivity and weak interpretability in the process of topic labeling,introducing ontology and association rules to construct frequent semantic patterns.[Method/process]Targeted tumor therapy patent data as the subject of study,firstly,by constructing an LDA2vec model to mine topics,then,considering the semantics and co-occurrence of topic words,the UMLS ontology is combined with the FPgrowth association rule to obtain the topic annotation with frequent semantic patterns,and finally the results are evaluated in comparison with the Canopy method.[Result/conclusion]The empirical results show that labeling based on frequent semantic patterns can generalize the meaning of topics from statistical and semantic perspectives and is more effective in topic annotation work.

关 键 词：主题标注 频繁语义模式 本体 关联规则 

分 类 号：R730.5[医药卫生—肿瘤] G353.1[医药卫生—临床医学]