摘要
伴随GIS深度和广度的不断增加,针对传统关系抽取模型存在复杂度高、耗时长的问题,本文提出了基于鲁棒主成分分析的空间GIS文本关系自主抽取。基于鲁棒主成分分析对空间采取降噪处理,实现低矩阵降维效果,根据不同的矩阵范数建立相应的空间关系。架构可度量的一维、二维、三维GIS空间,结合文本关系的自相似性,分析自主抽取过程;使用准确率、召回率、FI值作为关系抽取结果的评价指标,结合空间GIS文本变化幅率以及得到的文本间距离关系;对非线性问题采取特征变换处理,通过约束函数得到最终GIS文本关系抽取方程。通过仿真实验与文献方法对比,表明所提方法在准确率、耗时、召回率方面均优于传统方法,具有更高的鲁棒性,可被广泛应用于今后关系抽取研究中。
With the increasing depth and breadth of GIS,the traditional relation extraction model has high complexity and time-consuming problems.This paper proposes the autonomous extraction of spatial GIS text relations based on robust principal component analysis.Based on robust principal component analysis,noise reduction is applied to the space to achieve a low matrix dimensionality reduction effect,and corresponding spatial relationships are established according to different matrix norms.Structure measurable one-dimensional,two-dimensional and three-dimensional GIS space,and analyze the self-extraction process based on the self-similarity of text relations.Use the accuracy rate,recall rate,and FI value as the evaluation index of the relationship extraction results,and combine the spatial GIS text change rate and the distance relationship between the texts.Feature conversion processing for non-linear problems,and the final GIS text relation extraction equation is obtained through the constraint function.The comparison between simulation experiments and literature methods shows that the proposed method is superior to traditional methods in terms of accuracy,time-consuming,and recall.It has higher robustness and can be widely used in future relationship extraction research.
作者
王冠
Wang Guan(Wuhan Research Institute of Posts and Telecommunications,Wuhan 430070,China)
出处
《科技通报》
2021年第11期61-64,共4页
Bulletin of Science and Technology
关键词
鲁棒主成分分析
文本关系
自主抽取
范数
robust principal component analysis
text relation
autonomous extraction
norm
