摘要
【目的/意义】用户在进行文档内信息搜寻时,根据自身需求和阅读深入程度的不同,呈现出两种不同的阅读行为——深阅读与浅阅读。自动识别用户的深浅阅读有利于改善当前对深、浅阅读区分存在主观性强、耗时耗力的问题,对于研究文档内信息搜寻用户的个性化认知机制,优化用户信息搜寻体验也提供了很好的帮助。【方法/过程】本文根据前人对深、浅阅读的研究,利用K-means聚类算法构建文档内信息搜寻用户深、浅阅读行为的自动识别模型,并用实验验证模型分类的准确度。【结果/结论】实验结果显示,深、浅阅读在注视点持续时长、眼跳距离、眼跳方向和相邻注视点中心纵坐标距离这四个特征上有很大的差异,同时经过专家验证,K-means聚类模型识别深、浅阅读总准确率片段数为84.95%,片段时长为94.32%,达到了自动、准确识别文档内信息搜寻用户的深、浅阅读行为的效果。
【Purpose/significance】When searches information in documents,the user will present two different reading behaviors--deep reading and shallow reading,according to their own needs and the depth of reading.This study is helpful to solve the problem that the differentiation of deep and shallow reading behavior is subjective,time-consuming and labor-intensive.Meanwhile it provides a good ideal for studying the user’s personalized cognitive mechanism and optimizing the user’s reading experience.【Method/process】This paper analyzes the different eye movements between deep and shallow readings based on previous research results,then use K-means clustering algorithm to establish model,which automatically identifies user’s deep and shallow reading behaviors.【Result/conclusion】The experimental result shows that the deep and shallow readings have great differences in the four characteristics:duration of the fixation point,the distance of the eye,the direction of the eye,and the distance between the centers of the adjacent fixation points.And after experts’verification,the total accuracy of the K-means model recognition is 84.95%for the deep and shallow reading,and the segment duration is94.32%,which is very good for automatically and accurately identifying the user’s reading behavior,who search information in documents.
作者
陆泉
刘婷
刘庆军
陈静
LU Quan;LIU Ting;LIU Qing-jun;CHEN Jing(School of Information Management,Wuhan University,Wuhan430072,China;Center for Studies of Information Resources of Wuhan University,Wuhan430072,China;School of Information Management,Central China Normal University,Wuhan430079,China)
出处
《情报科学》
CSSCI
北大核心
2019年第10期126-132,139,共8页
Information Science
基金
教育部人文社会科学重点研究基地重大项目“大数据资源的挖掘与服务研究——面向医疗健康领域”(17JJD870002)
教育部人文社会科学研究规划基金项目“基于前景理论的信息搜索过程建模与预测研究”(18YJA870002)
关键词
深阅读
浅阅读
眼动追踪
信息搜寻
K-MEANS聚类
自动识别
deep reading
shallow reading
eye tracking
information seeking
K-means clustering
automatic recognition