基于语义特征和TextRank算法的科研成果论文中文文本关键词提取方法

1.山东思极科技有限公司,济南 250001;2.国网山东省电力公司,济南 250001

语义特征;TextRank算法;科研成果论文;中文文本;关键词提取;卷积神经网络

A Keyword Extraction Method for Chinese Text of Scientific Research Papers Based on Semantic Features and TextRank Algorithm
ZHANG Shichao1,WANG Jianbin2,MENG Hao2

1.Shandong SGIT Technology Co., Ltd., Jinan 250001, China;2.State Grid Shandong Electric Power Company , Jinan 250001, China

Semantic feature; TextRank algorithm; Scientific research paper; Chinese text; Keyword extraction;Convolutional neural network

DOI: 10.13512/j.hndz.2025.09.16

备注

为准确提取科研成果论文中文文本关键词,并准确排列,研究基于语义特征和TextRank算法的科研成果论文中文文本关键词提取方法。基于语义特征的科研成果论文中文文本候选关键词筛选方法,在Word2Vec工具中,将中文文本转换为词向量,作为论文中文文本语义特征;将语义特征输入卷积神经网络中,以分类的方式,提取属于候选关键词类型的语义特征,将其所属文本词语作为候选关键词;通过基于TextRank算法的科研成果论文中文文本关键词提取方法,在候选关键词中,以候选关键词的平均信息熵、词性、位置三种特征,为关键词提取指标,构建提取关键词的图模型,运算候选关键词综合权重,以从大到小的方式排列候选关键词,将排名靠前的候选关键词,作为最终提取的关键词,完成科研成果论文中文文本关键词提取。经测试,此方法可提高科研成果论文中文文本关键词提取精度、提高关键词排名准确性。
To accurately extract and arrange keywords from the Chinese text of scientific research papers, a keyword extraction method for Chinese text of scientific research papers based on semantic features and the TextRank algorithm was studied. A semantic feature-based method for selecting candidate keywords from Chinese text of scientific research papers was used. In the Word2Vec tool,the Chinese text was converted into a word vector as the semantic features of the Chinese text of the paper. The semantic features were input into convolutional neural networks, and the semantic features belonging to candidate keyword types were extracted through classification. The text words they belong to were used as candidate keywords. By using the TextRank algorithm-based keyword extraction method for Chinese text of scientific research papers, a graph model for extracting keywords was constructed by using the average information entropy, part of speech, and position of the candidate keywords as the keyword extraction indicators. The comprehensive weights of the candidate keywords were calculated,and the candidate keywords were arranged in descending order. The top-ranked candidate keywords were used as the final extracted keywords to complete keyword extraction from Chinese text of scientific research papers. The tests show that this method can improve the accuracy of keyword extraction and keyword ranking in Chinese text of scientific research papers.
·