情报杂志2025，Vol.44Issue(10)：163-171,9.DOI:10.3969/j.issn.1002-1965.2025.10.020
融合多特征注意力增强的关键词提取模型
The Keyword Extraction Model Enhanced by Multi-Feature Attention Fusion
摘要
Abstract
[Research purpose]Unsupervised keyphrase extraction does not require manually labeled data,making it widely applicable in the field of information processing.It holds significant research and practical value.However,existing embedding-based unsupervised models suffer from word vector anisotropy and semantic redundancy,leading to suboptimal keyword extraction performance.This study aims to solve these problems and improve the accuracy of keyword extraction and enhance the semantic representation ability.[Research method]This study proposes a Multi-feature Enhanced Attention Keyphrase Extraction Model(MEARank).The pre-trained model is used to obtain vector embeddings of phrases and documents,and the importance score of candidate phrases is calculated by combining the correlations between position features,phrase-document similarity,and self-attention graph semantic enhancement.An improved seman-tic similarity calculation function is also proposed to alleviate semantic repetition.[Research result/conclusion]The model was evaluated on three datasets,and the experimental results showed that the model achieved good experimental results,effectively improving the prob-lem of semantic repetition and having good application value.关键词
关键词提取/预训练模型/语义增强/注意力机制Key words
keyword extraction/pre-trained model/semantic enhancement/attention mechanism分类
信息技术与安全科学引用本文复制引用
魏聪运,张海军..融合多特征注意力增强的关键词提取模型[J].情报杂志,2025,44(10):163-171,9.基金项目
新疆维吾尔自治区自然科学基金项目"基于多特征融合的学生学习状态检测技术研究"(编号:2022D01A226) （编号:2022D01A226）
新疆维吾尔自治区重点研发计划项目"新疆专业技术人才数据治理与安全保障关键技术研发"(编号:2022B01007-1)研究成果. （编号:2022B01007-1）
