文献类型：期刊文献

中文题名：基于GloVe词向量的“技术——应用”发现研究

英文题名：Research on “Application-Technology” Discovery Based on GloVe Word Representation

作者：吉久明[1];施陈炜[1];李楠[1];康健[1]

机构：[1]华东理工大学科技信息研究所,上海200237

年份：2019

卷号：39

期号：4

起止页码：13

中文期刊名：现代情报

外文期刊名：Journal of Modern Information

收录：CSTPCD;;国家哲学社会科学学术期刊数据库;CSSCI:【CSSCI2019_2020】；

基金：2015年度教育部人文社会科学研究青年项目"面向语义出版的富语义模型构建与应用研究"(项目编号:15YJC870014)

语种：中文

中文关键词：词向量;GloVe模型;文本聚类;共现分析

外文关键词：word vector;GloVe model;text clustering;co-occurrence analysis

摘要：[目的/意义]本文探讨挖掘技术和应用间的关联关系,便于把握技术的目标应用场景和了解应用需求的可选技术方案,为技术转化提供基础。[方法/过程]首先,建立基于改进GloVe词向量的"技术——应用"发现模型,提出改进词向量的算法;然后,以机器学习技术为例,获取该领域学术论文文摘数据,预处理后建立训练文本集合,再使用改进GloVe模型训练获得词向量;最后,对词向量聚类获得"技术"和"应用"类词汇集,通过共现获得两者之间的关联并分析。[结果/结论]实验结果表明,本文改进GloVe模型后训练的词向量性能获得提升,聚类效果好,最终实现"技术——应用"的关联分析。
[Purpose/significance]The purpose of this paper was to mine the relationship between applications and technologies,and to know applications of technologies and optional technologies of applications,and provide the base for technology transformation.[Method/Process]Firstly,established an"application-technology"discovery model based on improved GloVe word vector and improved the word vector training algorithm.Then,collected academic literature abstracts in the field of machine learning,built the training text set after preprocessing,and used the improved GloVe model to obtain the word vector;finally,clustered the word vector to obtain the"application"and"technology"vocabulary sets,and acquired the relationship between them to analyze the cooccurence.[Result/Conclusion]The experimental results showed that the performance of the word vector obtained by the improved GloVe model was better,the clustering effect was good,and the correlation analysis of"application-technology"was finally realized.