基于SQA系统的社会化搜索答案质量评价指标构建

The Construction of Answer Quality Evaluation Index for Social Search: An Empirical Study on Social Q&A System

作为社会化搜索理念框架的一个重要组成部分,社会化问答系统(SQA)对Web 2.0用户的信息搜寻与知识获取方式产生了深远影响。在新一代互联网用户生成内容(UGC)环境下,虽然有关信息质量及其评价标准与维度的研究已取得一定程度的进展,但国内目前对SQA系统答案质量要素、SQA信息质量评价指标的研究仍缺乏足够的深度。本文以现有相关研究为基础,在社会化搜索的研究背景下,借助"百度知道"用户对答案质量感知的问卷,综合采用专家访谈、探索性因子分析和验证性因子分析方法,构建了一个基于SQA系统的社会化搜索答案质量评价模型SQA-IQ4Ⅲ,该模型包括内容质量、情境质量、来源质量和情感质量四个维度,共有18项关键性指标要素。研究结论将进一步扩展图书情报学和网络信息资源管理的研究领域,并对SQA用户获取高质量答案以及UGC产品/服务设计产生实际的帮助。

SQA （ Social Question and Answering） is one of the typical applications of Social Search in Web 2. 0 era. SQA sites take advantage of the wisdom of crowds and produce a mass of questions and answers every day. In a new generation of Internet UGC （ User-Generated Content） environment, SQA has a profound effect on the way of information seeking by users. To better understand the successful implementation of SQA and its underlying challenges and issues, researchers from a variety of fields of LIS （ Library ＆ Information Science）, such as information search, information behavior, and human computer interaction have devoted to identify the user＇s motivation and behavior in SQA for the past few years. Those user-oriented studies may focus on the cognitive, social and psychological aspects of answer adoption and diffusion of SQA. However, there is still a lack of comprehensive and systematic understanding toward the topic of investigating the assessment or the indicators of answer quality for SQA. On the basis of the existing studies and the research context of social search, this paper aims to develop an evaluation index of answer quality of SQA from the user perspective. Baidu Knows （zhidao.baidu.com）, which is the largest Chinese community of SQA website, is used to conduct the empirical study. A mixedmethod research with scholar interview and survey was employed to investigate the dominant factors of answer quality of SQA by Exploratory Factor Analysis （EFA） and Confirmative Factor Analysis （CFA）. First, in order to improve the evaluation index and adjust the original indicators, before the questionnaire survey and analysis, scholar interviews were assumed to detect indicators that were ignored or not easily understood in this paper. Five scholars with deep knowledge and rich research experience in the field of information quality were selected as our interview objects. The primary purpose of conducting the interview was to focus on the rationality and integrity of the index, which helps us in evaluating whether the indicators proposed indeed measured the answer quality constructs in SQA context. Then, we employed users of Baidu Knows as the investigative object to collect data, and issued an e-questionnaire （the answer quality survey of Baidu Knows） on the website named Sojump （www.sojump.com） to complete the survey. The questionnaire mainly consisted of two parts. The first part was related to the demographic characteristic of interviewees, the average time they used to ask for an answer, as well as the themes they were interested in. The second part was the measurement about the level of answer quality that users perceived from the usage of and the interaction with SQA. Afterwards, a principal component analysis was conducted to test the preliminary survey data, and the index model was revised for the first time. Finally, an EFA using SPSS 18. 0 and a CFA using AMOS 17.0 were synthetically employed to test the formal survey data, and the indicators of the model were amended for the second time. The research findings show that, the evaluation index of answer quality of SQA includes four dimensions, i.e., Content Quality, Context Quality, Source Quality and Emotional Quality with a total of 18 key indicators in detail. Theoretically, this study is significant to expand the research fields of LIS, web information resources management, and quality management. Practically, the research may yield some implications for the interaction design of SQA systems and UGC applications. In the future work, in order to discuss information quality perceived and evaluated by users of different groups for SQA, we will conduct a classified statistic to explore the average time that an asker spent in seeking for information, and topics concerned by interviewees in. Meanwhile, we will give full consideration to the reciprocal effect of four dimensions in this evaluation index, and add some regulative factors to a more detailed exploration. 2 figs. 9 tabs. 40 refs.