图书情报工作 ›› 2019, Vol. 63 ›› Issue (11): 118-130.DOI: 10.13266/j.issn.0252-3116.2019.11.013

• 知识组织 • 上一篇    下一篇

社会化问答社区用户生成答案质量自动化评价研究——以“知乎”为例

郭顺利1, 张向先2, 陶兴2, 张莉曼2   

  1. 1. 曲阜师范大学传媒学院 日照 276826;
    2. 吉林大学管理学院 长春 130022
  • 收稿日期:2018-08-05 修回日期:2018-11-18 出版日期:2019-06-05 发布日期:2019-06-05
  • 作者简介:郭顺利(ORCID:0000-0002-3155-9937),讲师,博士,E-mail:guosl777@163.com;张向先(ORCID:0000-0003-3186-2677),教授,博士,博士生导师;陶兴(ORCID:0000-0003-0480-4201),博士研究生;张莉曼(ORCID:0000-0002-0770-3708),博士研究生。
  • 基金资助:
    本文系吉林大学研究生创新基金资助项目"移动商务用户在线评论的隐式意见挖掘及可视化研究"(项目编号:2017082)研究成果之一。

Research on Automated Evaluation of User Generated Answer Quality in Social Question and Answer Community——Taking “Zhihu” as an Example

Guo Shunli1, Zhang Xiangxian2, Tao Xing2, Zhang Liman2   

  1. 1. Media College, Qufu Normal University, Rizhao 276826;
    2. Management School Jilin University, Changchun 130022
  • Received:2018-08-05 Revised:2018-11-18 Online:2019-06-05 Published:2019-06-05

摘要: [目的/意义]旨在构建社会化问答社区用户生成答案质量评价指标体系,实现面向用户需求的答案质量自动化评价和筛选,提高社会化问答社区知识服务质量。[方法/过程]引入社会情感特征和用户特征,运用因子分析和结构方程实证构建用户生成答案质量评价指标体系。基于GA-BP神经网络模型设计答案质量自动化评价方法。最后,选取知乎网站数据对用户生成答案质量评价指标体系和自动化评价方法进行应用研究。[结果/结论]构建包含答案文本特征、回答者特征、时效特征、用户特征、社会情感特征5个维度的评价指标体系。实验分析发现基于GA-BP神经网络的答案质量自动化评价方法相比于其他方法准确率较高、平均误差低,具有可行性和有效性,能够进一步应用和推广实践。

关键词: 社会化问答社区, 用户生成答案, 质量评价, 用户需求

Abstract: [Purpose/significance] The paper aims to build the social QA community users to generate the quality evaluation index system, achieve automatic evaluation and selection of answers to user needs, and improve the quality of the community QA community service.[Method/process] The introduction of social emotional features and user characteristics, and factor analysis and structural equation analysis are used to build an index system for evaluating the quality of user generated answers. Then, based on the GA-BP neural network model, the automatic evaluation method of the answer quality is designed. The application of the quality evaluation index system and automatic evaluation method of user generated answers is studied.[Result/conclusion] The evaluation index system consists of 5 dimensions, including the characteristics of the answer text, the characteristics of the respondent, the timeliness, the user characteristics and the social emotional characteristics. The experimental analysis shows that the method of automatic evaluation of the answer quality based on GA-BP neural network is more accurate and lower than other methods. It is feasible and effective, and can be further applied and popularized.

Key words: social question and answer community, user generated answers, quality evaluation, user demand

中图分类号: