【作者】 成颖；

【导师】 孙建军；

【作者基本信息】 南京大学 ， 情报学， 2011， 博士

【摘要】 相关性是情报学基本问题的观点得到了国内外图书情报学界的普遍认同。国外的相关性研究无论在理论还是实证方面都取得了丰硕的成果,相关成果可参见Saracevic(1975)、Mizzaro (1997)、Schamber(1994)以及Saracevic (2007)等完成的综述。相关性研究虽然也得到了国内学界的重视,但是只能检索到为数不多的国外研究成果的综述与介绍。鉴于此,本文开展了基于本土的面向用户的相关性判据研究以及相关性判据应用研究。本研究的主要成果与结论有：(1)相关性判据集。本文以南京大学信息管理系4个年级以及南京大学教育科学系1个年级的相关性判据文本数据为信息源,采用内容分析法解析出了包括传播特征,内容,情境,使用,系统特征,愉悦,质量,总体以及文献特征等9类相关性判据。通过频次分析发现,文献特征、质量判据以及内容判据、总体和使用占据了相关性判据的主要位置,而愉悦、情境、传播特征以及系统特征则相对没有那么重要。与Schamber(1991)以及Barry(1993)等前人研究相比,本研究拓展了面向用户的相关性判据集,提炼出了包括文献总体、文献使用以及传播特征等三个新的相关性判据类别。(2)影响信息用户相关性判断的文献特征。通过对相关性判据文本数据的内容分析,解析出了包括参考文献、出版单位、出版时间、关键词、基金、篇幅、全文、题名、文献格式、文摘、语种、文献类型、来源期刊、作者以及机构等在内的15个影响信息用户相关性判断的文献特征。(3)任务复杂性以及性别影响相关性判据的选择。研究结果显示,用户在面对不同复杂性的任务时,在文献内容、文献使用、文献特征、作者以及文献类型等相关性判据的选择方面存在差异；而性别维度仅影响作者、文献类型以及文献总体层面的相关性判据选择。由于数据量的原因,本文没能证实任务复杂性以及性别对信息用户在选择来源期刊、机构、系统特征、传播特征以及愉悦感知等5个类别相关性判据时的影响。(4)基于相关性判据、价值增值模型、TEDS模型以及信息系统成功模型构建了面向相关性判据的学术信息检索系统成功模型,并证实了模型的有效性。根据模型设计了问卷,问卷共发放1114份,回收1054份,其中有效问卷929份。通过对数据的结构方程分析证实：①系统因子1和系统因子2对选择性存在正负不同方向的影响,路径系数分别为0.59和-0.39。②系统因子2、系统因子3以及选择性对系统效能存在正性的影响,路径系数分别是0.11、0.15和0.50。③系统因子1和灵活性对自适应性存在正性的影响,路径系数分别达到0.14和0.70。④系统因子5、灵活性、时间认知以及可靠性对系统性能存在正性的影响,路径系数分别为0.20,0.29,0.10以及0.38。⑤美感和娱乐体验对情感认知存在正性的影响,路径系数分别为0.54,0.27。⑥系统因子4和系统因子6对易用认知存在正性的影响,路径系数分别为0.23,0.21。⑦完整性、实时性、权威性、有效性对于信息质量存在正性的影响,路径系数分别为0.26,0.26,0.38以及0.25。⑧系统因子1、灵活性、隐私以及自适应性对服务质量存在正性的影响,路径系数分别为0.10,0.15,0.10,0.52。⑨效能认知、性能认知、选择性、易用认知、情感认知以及自适应性对系统质量存在正性的影响,路径系数分别为0.21,0.30,0.29,0.08,0.32,0.23。⑩系统质量、信息质量对满意度存在正性的影响,路径系数分别为0.56,0.29。最后,结构方程分析的结果表明,信息质量、系统质量以及满意度对使用意图存在正性的影响,路径系数分别为0.35,0.55以及0.11,而服务质量对于使用意图存在负性的影响,路径系数为-0.16。更多还原

【Abstract】 It has been widely recognized that relevance is one of the basic problem in library and information science. Foreign studies have made abundant achievements theoretically and practically, for more information please consult Saracevic(1975), Mizzaro(1997),Schamber(1994) and Saracevic (2007),et al. Though more and more native scholars are paying attention to Relevance, only some introductions and overviews of foreign research progress can be retrieved from CNKI. In view of these, a serial of user-oriented research on relevance criteria and their application in a native atmosphere are carried on in this article. Primary results and conclusions are listed as follows:(1) Provision of a set of relevance criteria. The paper selected 4 grades in the information management department and 1 grade in education science of Nanjing University as data resources, and provided 9 kinds of relevance criteria including broadcasting features, content, context, use, system features, pleasure, quality, collection and document feature. By analyzing the frequencies of these relevance criteria, it has been found that document feature, quality, content, collection and use accounted for the main location, while pleasure, context, broadcasting feature are not so important. Compared to Schamber’s (1991) and Barry’s(1993), the paper extends the set of relevance criteria, and parse out three new relevance criteria including document collection, document use and broadcasting feature.(2) Document features which affect the user’s relevance judge. By the content analysis to the relevance criteria text, the author parse out 15 document features including references, publishers, publishing time, keyword, funds, space, text, title, document format, abstracts, language, document type, source journals, authors and institutions, which have impacts on user’s relevance judge.(3) Impact of task complexity and gender on relevance criteria’s selection. Research show that different user have differences in the selection of relevance criteria including document content, document use, document features, author, and document type when are faced to tasks with different complexion.(4) Construction of academic information retrieval system success model oriented to relevance criteria based on relevance criteria, value added model, TEDS model and ISSM as well as validation of its validity. A questionnaire is design based on the model,1114 questionnaires were distributed and 1054 were returned of which 929 questionnaires were valid. By SEM method, we prove that:①System fact 1 and system fact 2 have positive and negative impact on selection, with the path coefficient 0.59 and-0.39 each.②System fact 2, system fact 3 and selection have positive impact on self-adaption, with the path coefficient 0.11,0.15 and0.50 each.③System fact 1 and flexility have positive impact on self-adaption, with the path coefficient 0.14 and 0.70 each.④System fact 5, flexility, perceived time and reliability have positive impact on system performance, with the path coefficient 0.20,0.29,0.10 and 0.70 each.⑤Beauty and play experience have positive impact on perceived emotion, with the path coefficient 0.54 and 0.27 each.⑥System fact 4 and system fact 6 have positive impact on perceived ease of use, with the path coefficient 0.23 and 0.21 each.⑦Completeness, timeliness, authority, effectiveness have positive impact on the quality of information, with the path coefficient 0.26,0.26,0.38 and 0.25 each.⑧system fact 1, flexility, privacy and self-adaptation have positive impact on the service quality. with the path coefficient 0.10,0.15,0.10, and 0.52 each.⑨Perceived efficiency, perceived performance, selectivity, perceived ease of use, perceived emotion, and self-adaptability have positive impact on the system quality, with the path coefficient 0.21,0.30,0.29,0.08,0.32 and 0.23 each.⑩System quality and information quality have positive impact on satisfaction, with the path coefficient 0.11,0.15 and 0.50 each. Finally, the SEM shows that information quality, system quality and satisfaction have positive impact on intention to use with the path coefficient 0.35, 0.55 and 0.11 each, while the service quality has negative impact on intention to use with the path coefficient-0.16更多还原