主题词法和自然语言法探测文献主题新颖性对比分析
|
摘要
目的:对比分析主题词法和自然语言法计算结果的一致性和差异性,探讨两种方法的优缺点以及与F1000推荐文献的关系。方法:定义医学主题词词对法的文档主题新颖度概念,给出计算公式进行计算并进行对比分析。结果:主题词法计算了该文献集401篇文献中已标引的346篇文献的文档主题新颖度,平均新颖度值为0.8423;自然语言法计算了该文献集全部401篇文献的文档主题新颖度,平均新颖度值为0.8713。74.28%的文献经两种方法计算得到的新颖度差值在0.1以下。结论:主题词法和自然语言法可从文本层面计算文档主题新颖度,两者各有优势。自然语言法在计算范围和最新发表的文献方面要略优于主题词法,主题词法在揭示文章主旨含义和准确度方面,优于自然语言法。根据相关性比较,主题词法和自然语言法在计算文档主题新颖度方面一定程度上具有相对等效的价值。新颖度值越高,主题词法和自然语言法计算出的文档新颖度值分区越一致。主题词法文档主题新颖度与F1000得分弱相关,说明主题词法新颖度准确性更接近专家同行评议。
Objective To comparatively analyze the consistency and difference calculated according to MeSH and natural language and to study the relationship of their advantages and disadvantages with the F1000 recommended papers.Methods The concept of literature topic novelty detected with MeSH pairs was defined. The formula for calculating the literature topic novelty was presented and comparatively analyzed. Results The topic novelty of the indexed 346 papers from the 401 papers and the 401 papers was calculated according to MeSH and natural language respectively. The average novelty of the indexed 346 papers and the 401 papers was 0.8243 and 0.8713 respectively.The difference-value of novelty in 74.28% papers calculated according to Me SH and natural language was < 0. 1.Conclusion MeSH and natural language can calculate the novelty of literature and have their advantages and disadvantages. Natural language is better than MeSH in calculating the novelty of recently published papers while Me SH is better than natural language in revealing the meaning and accuracy of papers. Correlation The value of Me SH and natural language is equivalent in calculating the literature topic novelty.The higher the novelty is,the more consistent the literature novelty value is. The literature topic novelty is weakly related with the F1000 score,indicating that the accuracy calculated according to Me SH is closer to that calculated according to peer review.
Objective To comparatively analyze the consistency and difference calculated according to MeSH and natural language and to study the relationship of their advantages and disadvantages with the F1000 recommended papers.Methods The concept of literature topic novelty detected with MeSH pairs was defined. The formula for calculating the literature topic novelty was presented and comparatively analyzed. Results The topic novelty of the indexed 346 papers from the 401 papers and the 401 papers was calculated according to MeSH and natural language respectively. The average novelty of the indexed 346 papers and the 401 papers was 0.8243 and 0.8713 respectively.The difference-value of novelty in 74.28% papers calculated according to Me SH and natural language was < 0. 1.Conclusion MeSH and natural language can calculate the novelty of literature and have their advantages and disadvantages. Natural language is better than MeSH in calculating the novelty of recently published papers while Me SH is better than natural language in revealing the meaning and accuracy of papers. Correlation The value of Me SH and natural language is equivalent in calculating the literature topic novelty.The higher the novelty is,the more consistent the literature novelty value is. The literature topic novelty is weakly related with the F1000 score,indicating that the accuracy calculated according to Me SH is closer to that calculated according to peer review.
引文
[1]邢美凤,过仕明.文本内容新颖性探测研究综述[J].情报科学,2011,29(7):1098-1103.
[2]邱均平.文献信息引证规律和引文分析法[J].情报理论与实践,2001,24(3):236-240.
[3]施聪莺,徐朝军,杨晓江.TFIDF算法研究综述[J].计算机应用,2009,29(6):167-170,180.
[4]徐爽.基于突发监测的全身炎症反应综合征治疗药物研究趋势分析[D].沈阳:中国医科大学,2010.
[5]杨建林,钱玲飞.基于关键词对逆文档频率的主题新颖度度量方法[J].情报理论与实践,2013,36(3):99-102.
[6]陈斯斯,董立平,许丹,等.医学文献主题新颖性探测方法对比分析[J].中华医学图书情报杂志,2018,27(2):20-25.
[7]许丹,徐爽,陈斯斯,等.基于自然语言词对法的文献主题新颖性探测研究[J].图书情报工作,2018,62(8):130-138.
[8]廖星,谢雁鸣.F1000:新兴医学论文评价和检索系统[J].中国中西医结合杂志,2012,32(5):701-703.
[9]王岩,张静仪,何晓阳.PubMed与Embase的主题标引及主题检索功能比较研究[J].中华医学图书情报杂志,2018,27(6):35-41.
[10]Welcome to Medical Subject Headings[EB/OL].[2018-11-28].https://www.nlm.nih.gov/mesh/meshhome.html.
[11]张阳德.microRNA的肿瘤表达研究及协同调控网络分析[D].长沙:中南大学,2014.
[12]刘春丽,何钦成.不同类型选择性计量指标评价论文相关性研究:基于Mendeley、F1000和Google Scholar三种学术社交网络工具[J].情报学报,2013,32(2):206-212.
[2]邱均平.文献信息引证规律和引文分析法[J].情报理论与实践,2001,24(3):236-240.
[3]施聪莺,徐朝军,杨晓江.TFIDF算法研究综述[J].计算机应用,2009,29(6):167-170,180.
[4]徐爽.基于突发监测的全身炎症反应综合征治疗药物研究趋势分析[D].沈阳:中国医科大学,2010.
[5]杨建林,钱玲飞.基于关键词对逆文档频率的主题新颖度度量方法[J].情报理论与实践,2013,36(3):99-102.
[6]陈斯斯,董立平,许丹,等.医学文献主题新颖性探测方法对比分析[J].中华医学图书情报杂志,2018,27(2):20-25.
[7]许丹,徐爽,陈斯斯,等.基于自然语言词对法的文献主题新颖性探测研究[J].图书情报工作,2018,62(8):130-138.
[8]廖星,谢雁鸣.F1000:新兴医学论文评价和检索系统[J].中国中西医结合杂志,2012,32(5):701-703.
[9]王岩,张静仪,何晓阳.PubMed与Embase的主题标引及主题检索功能比较研究[J].中华医学图书情报杂志,2018,27(6):35-41.
[10]Welcome to Medical Subject Headings[EB/OL].[2018-11-28].https://www.nlm.nih.gov/mesh/meshhome.html.
[11]张阳德.microRNA的肿瘤表达研究及协同调控网络分析[D].长沙:中南大学,2014.
[12]刘春丽,何钦成.不同类型选择性计量指标评价论文相关性研究:基于Mendeley、F1000和Google Scholar三种学术社交网络工具[J].情报学报,2013,32(2):206-212.