摘要
用户对网页内容的评估结果,是了解用户信息需求的抓手。信息服务向智能化、个性化方向发展,要求挖掘用户搜索行为规律,自动识别相关网页。眼动数据直接反映阅读状态,具有与心理活动联系密切的特点,国(境)外将其作为相关性反馈数据,经过十余年研究积累了一些成果,但简单指标效果不稳定、复合指标计算烦琐,目前尚无自然搜索环境下可投入使用的眼动指标。文章基于既有指标导出新复合指标,并以疾病自诊为实验背景进行了验证。结果显示,跳视/注视比(ScFi)反馈效果良好,配合连贯阅读注视点数这一指标使用,判别正确率可达95.6%,当前样本情况下,两个指标阈值分别为0.187、30。
User’s evaluation results of web pages are the main way to understand the user’s information needs.To realize intelligent and personalized development of information service,it requires mining users’patterns of searching behaviors and identifying relevant web pages automatically.Eye movement data reflect the state of personal reading and are closely related to psychological activities,which are used as the feedback of relevance at abroad.In the past ten years,some achievements have been made in the area of eye movement indexes for page relevance,however,existing simple indicators are unstable and composite indicators are too complex for practice.So far,there are no eye movement indexes that can be put into use.The paper tries to explore new composite indexes according to existing indexes,and verifies the new indexes through the experiment of self-diagnosis search.The results show that the feedback of the ratio of saccade time and fixation time(ScFi)is sound,and combining with the index of coherent reading fixation points,the accuracy of prediction reaches 95.6%.The threshold value of ScFi and coherent fixation points is 0.187 and 30.
出处
《情报理论与实践》
CSSCI
北大核心
2019年第2期164-168,共5页
Information Studies:Theory & Application