基于自然语言语义感知的敏感信息识别与分析算法

打开文本图片集

引用格式：.基于自然语言语义感知的敏感信息识别与分析算法[J].现代电子技术,2025,48(18)：17-21.

关键词：自然语言处理；敏感信息；语义感知；特征提取；信息识别；长短期记忆模型；Transformer中图分类号：TN919.8-34；TP391.1 文献标识码：A 文章编号：1004-373X（2025）18-0017-05

Sensitive information recognition and analysis algorithm based on natural language semantic perception

ZHANG Xiaoyan (Xi’anUniversityofFinanceandEconomics,Xi'an71oiOo,China)

Abstract:Acomprehensiveprocessing algorithmbasedon long short-termmemory (LSTM）andconvolutional neural network(CNN）isproposedtoaddress thegrowing problemofsensitiveinformationanalysisinnaturallanguage.Inthis algorithm，keyvocabularyisdividedafterinputdataareencoded，allvocabularyissummarizedstatisticalyndssitie vocabulary listisconstructed.Then,thesortingalgorithmisusedtosortandstorethesensitivewordsacordingtotheir frequencyofoccurrence.TherepresentationvectorsareextractedfromsensitiveinformationtablesbymeansofBiLSTM,soasto characterizethefeaturedegreeofsensitiveinformationinthesample.BasedonthematrixcalculationmethodintheTransformer structure, the Q query vector is set as the extracted text representation vector,and the V value matrix isset as the relevant parametervaluesoftheextractedsensitiveinformationtocalculatetheatentionvalue.Furthermore,theatentionfeaturesbased onfeatureposition expresionareintroducedtocomprehensivelyimprovetherecognitionacuracyof thealgorithm.The experimental results show that the recognition accuracy of the proposed algorithm is increased by about 2.319% compared to the optimal comparisonalgorithm.Theablation experimentalcomparisonresultsshowthattherecogitionaccuracyof theimproved algorithmisincreasedby 6.941% compared to the original algorithm.

Keywords:natural language process；sensitiveinformation；semantic perception；feature extraction;information identification;long short-term memory model; Transformer

0 引言

随着智能终端和信息技术的飞速发展，人们接触与传播信息的方式日益多样化，网络社交媒体的信息传播速度和范围达到了更高的水平。（剩余9563字）