【作者】 王丹；

【导师】 陈静；

【作者基本信息】 华中师范大学 ， 管理科学与工程， 2019， 硕士

【摘要】 信息技术的快速发展催生了大量的网络资源,科研领域也不例外。海量的学术资源使得研究人员需要花费更多的时间和精力去获得满足自己需求的信息,科研论文的主动推荐解决了急剧增长数字化文献带来的信息过载问题,帮助减轻用户阅读非相关文献带来的时间成本负担。目前论文个性化推荐主要根据用户的研究兴趣、社会网络关系等构建用户画像,然后与论文进行相似度匹配得到推荐论文。然而依靠用户研究兴趣或社会属性这类外部特征构建的用户画像存在局限,内在心理感知视角的缺失使得画像并不全面,例如根据关键词只能将兴趣相似的一类人识别出来,并不能识别出兴趣相似知识水平不一致的用户。有学者指出不同研究经验的用户对推荐文章的认知是不同的,用户的信息行为是一个认知活动,复杂的认知过程决定了人们对于学科知识有着不同的组织方式及掌握程度。因此从心理认知层面识别出用户的认知偏好,针对不同水平的用户推荐与他们知识水平一致的论文,能够实现科研论文学术资源更为精准的推荐,帮助用户获得满足心理需求的论文。本文在梳理目前国内外学者在科研论文推荐领域的相关研究时,首先针对科研用户画像未衡量用户认知差异的问题,介绍了认知心理学中的用户认知特征类型;接着从科研论文质量值计算方法以及推荐结果评价方法及指标等方面对论文推荐研究进行了阐述。在理论梳理的基础上,文章提出了基于用户认知的科研论文推荐模型,招募不同经验水平的学生作为受试者进行了论文推荐实验,通过实验验证了考虑用户认知结构及能力在结构匹配、能力匹配、需求度、整体满意度四个指标上都能很好地满足高级组被试的需要;而对于低级和中级组而言,考虑认知结构及能力的推荐方法能够得到在部分指标上获得最优的推荐效果。最后,本文对整个基于用户认知的科研论文推荐研究进行了总结,指出了研究和实验中存在的不足以及未来研究的方向。更多还原

【Abstract】 The rapid development of information technology has generated a large number of network resources,and academic field is no exception.The overabundance of academic information has forced scholars to spend more time and resources in searching for information relevant to their needs.The emergence of recommendation technology for academic paper helps to solve the information overload problem,and relieve the burden of time wasted on reading irrelevant articles for researchers.Existing literature on personalized academic paper recommendation mainly build user profile based on users’research interest or social connections,and then match with papers to get the recommended list.However,user profile built based on these external features(ie:research interest,social attributes)is limited and incomprehensive with the lack of inner perception.For example,keywords matching can solely identify users who express similar interests,rather than users with similar interests while differ in knowledge level.It has been proved that users of varying experience levels have different perception to the recommended paper.User information behavior can be seen as a cognition activity,the complicated cognition process decides users’ different ways in organizing and mastering knowledge.Therefore,identifying users’ predilection from the level of psychological cognition,recommending papers that are consistent with their knowledge level,the academic paper recommendation system can achieve more accurate recommendation result,and help users to obtain papers that meet psychological needs.When reviewing the relevant researches in the field of academic paper recommendation,this paper firstly introduces the types of user cognitive characteristics in cognitive psychology,with the limitation of ignorance of user cognition in current user profile.Then the paper describes the calculation method of academic paper quality,the evaluation methods and indicators of the recommendation results.On the basis of these theoretical support,this article proposes a paper recommendation model based on user cognition,and recruited students with different experience levels as subjects to conduct experiment.The experiment result verifies that considering the user’s cognitive structure and ability can meet the needs of the high ability group on the indicators of structure matching,ability matching,demand and overall satisfaction.For the lower and middle ability groups,the kind of recommended method can obtain the best recommendation effect on some indicators.Finally,this paper summarizes the research of academic paper recommendation based on user cognition,points out the weakness in research and experiment and the future research prospect.更多还原