学术文本的结构功能识别--在关键词自动抽取中的应用  

Structure-Function Recognition of Academic Text——Application in Automatic Keywords Extraction

作　　者：方龙[1] 李信[1] 黄永[1] 陆伟[1] 

机构地区：[1]武汉大学信息管理学院,信息检索与知识挖掘实验所,武汉430072

出　　处：《情报学报》2017年第6期599-605,共7页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金面上项目“面向词汇功能的学术文本语义识别与知识图谱构建”(71473183);国家自然科学基金面上项目“基于多语义信息融合的学术文献引文推荐研究”(71673211)

摘　　要：当前的关键词自动提取研究大多基于候选词的词频、文档频率等统计信息,往往忽略了侯选词所在的学术文本的内在结构,导致关键词提取的效果不佳。本文将学术文本看作是5个结构功能域的集合,提出了融合学术文本结构功能特征的多特征组合提取方法,并利用学术文本的章节标题对其结构功能进行识别,然后通过SVM二分类和LambdaMART学习排序算法分别在计算机语言学领域的文献集上进行了实现。实验结果表明,本文提出的组合特征方法相比基准特征在关键词提取的效果上取得了较大的提升,尤其在分类实验中准确率的相对提升上达到10.75%,证明了学术文本结构功能特征在关键词自动提取上的重要性。The current approach for automatic extraction of keywords is mostly based on the frequency and docu- mentation frequency of the alternative words, which are ignored in the inner structure of academic text and leads to bad work of keywords extraction. In this article, we considered an academic text as a collection of five struc- ture-function models, proposed a mixed feature method with academic text structure features, and implemented it in the computer science documents by a classification technique called support vector machine （SVM） and a ranking model named LamdbaMART. The results show that the methods we put forward are more effective than the baseline model with base features and a 10.75% relative improvement has been observed on the precision in classification of experiment, which proves that the academic text structure is important for automatic keywords extraction.

关 键 词：结构功能 关键词提取 学术文本 支持向量机 学习排序 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]