基于引文的科技文献主题抽取研究

发布时间：2018-01-02 12:05

本文关键词：基于引文的科技文献主题抽取研究　出处：《情报理论与实践》2017年06期 　论文类型：期刊论文

【摘要】：[目的/意义]主题抽取的效果对于信息检索、自动标引、自然语言处理具有重要的价值,提高主题抽取的效果,既能改善检索系统主题检索准确性,又能够帮助学者更加高效地了解文献的主要思想。文章研究并探讨了从引用内容中抽取文献主题的有效性。[方法/过程]选取自然语言处理领域文献为研究对象,利用文献之间的引用与被引用关系抽取引用内容,进行分词并计算权重;将引用内容、全文抽取的候选词进行专家打分评价效果并将其与关键词对比,使用F值评价两种语料库抽取候选词的优劣。[结果/结论]通过专家打分及计算F值,发现引用内容在抽取候选词方面具有明显优势。
[Abstract]:[Objective / meaning] the effect of topic extraction is of great value to information retrieval, automatic indexing and natural language processing. It can improve the accuracy of topic retrieval in retrieval system. It can also help scholars to understand the main ideas of literature more efficiently. [Methods: the natural language processing literature was chosen as the research object. The citation content was extracted from the citation and the citation relation between the references, and the word segmentation was carried out and the weight was calculated. The candidate words extracted in full text are evaluated by experts and compared with the key words, and the advantages and disadvantages of the two kinds of corpus extraction are evaluated by using F value. [Results / conclusion] by scoring and calculating F value of experts, it is found that citation has obvious advantages in extracting candidate words.
【作者单位】：华中师范大学信息管理学院;
【基金】：国家社会科学基金项目“基于语义理解的多模态信息融合与集成服务研究”(项目编号:11CTQ017) 中央高校基本科研业务费专项资金项目“社交化问答平台信息质量感知模型及推荐算法研究”(项目编号:CCNU16A02035)的成果
【分类号】：G353.1
【正文快照】： 主题是对文献中心思想的概括,有了主题就基本了解了文献阐述的主要内容。从检索系统来看,篇名、关键词、主题、摘要、全文都是文献的检索入口,但是这些入口对于文献主题的反映准确性参差不齐,降低了检索人员效率。引用内容通常是学者对文献观点、主张高度凝练的概括,具有观点

【参考文献】