面向航空领域HowNet构建的术语内部动态角色标注  

EVENTROLE LABELING IN TERMS ORIENTED TO HOWNET CONSTRUCTION IN AVIATION FIELD

作　　者：赵超丽 王裴岩[1] 蔡东风[1] Zhao Chaoli;Wang Peiyan;Cai Dongfeng(Human-computer Intelligence Research Center,Shenyang Aerospace University,Shenyang 110136,Liaoning,China)

机构地区：[1]沈阳航空航天大学人机智能研究中心,辽宁沈阳110136

出　　处：《计算机应用与软件》2021年第3期163-168,209,共7页Computer Applications and Software

基　　金：教育部人文社会青年科学研究基金项目(17YJC740087)。

摘　　要：术语内部动态角色标注是航空领域HowNet构建的关键环节,其直接影响航空领域HowNet的规模与质量。针对动态角色种类多造成标注困难的问题,提出一种基于KNN的术语内部动态角色标注方法。通过对术语内部词语DEF项的分析进行样本预选择,并在最近邻样本选择阶段融合基于DEF的语义相似性及基于词向量的语境分布相似性。实验结果表明,1-Best、3-Best和7-Best的准确率分别为67.57%、86.00%和94.17%,平均倒数排名MRR为0.7764,优于现有的研究结果。EventRole labeling within terminology is a key link in the construction of HowNet in aviation field.Its labeling method directly affects the scale and quality of HowNet in aviation field.Aiming at the difficulty of EventRole labeling caused by many kinds of EventRoles,we propose a term internal EventRole labeling method based on KNN classification algorithm.The method used the DEF terms of the internal two words to pre-extract the training sample data of the KNN algorithm,and in the nearest neighbor sample selection stage,the semantic similarity based on DEF and the similarity of context distribution based on word embedding were combined.The experimental results show that,the accuracy rates of 1-Best,3-Best,and 7-Best are up to 67.57%,86.00%,and 94.17%,and the average reciprocal ranking is up to 0.7764,which are better than the existing work results.

关 键 词：航空术语 动态角色 HOWNET KNN分类算法 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]