摘要
【目的/意义】基于弹幕的用户画像描述了用户的行为模式,有助于弹幕视频平台理解用户的需求与偏好,针对性地改善平台的内容与服务,以增强用户参与度与忠诚度。【方法/过程】以B站教学类视频为例,采集弹幕与视频数据,计算用户发送弹幕的内容特征与行为特征,根据特征对用户聚类得到用户画像,并探讨各个特征的内在关系。【结果/结论】研究结果表明:用户从行为特征角度可以分为交互性优先学习者、一般学习者、质量优先学习者与完整性优先学习者;从内容特征角度可以分为签到、表达学习感受、表达情感与讨论课程知识四种内容偏好。用户发送弹幕的频率越高,真实花费在观看视频上的时间就越少。完整性优先取向的用户更倾向于讨论课程知识,而不倾向于表达学习感受。【创新/局限】创新性地融合了用户的弹幕文本内容与时间信息,实现了更为全面的用户画像。后续研究可继续探讨不同内容主题的视频中用户画像的区别与联系。
【Purpose/significance】User portraits based on the danmaku describe the user's behavior pattern,which helps the video platform to understand the needs and preferences of users,and to improve the content and services of the platform to enhance user participation and loyalty.【Method/process】Take the teaching video of Bilibili as an example,collect danmaku and video data,calculate the content and behavior characteristics of the danmaku sent by users,cluster users according to the features to obtain user portraits,and explore the internal relationship of each feature.【Result/conclusion】From the perspective of behavior characteristics,users can be divided into interactive first learners,general learners,quality first learners and integrity first learners;from the perspective of content characteristics,they can be divided into four types:sign-in,learning feelings expression,emotion expression and course knowledge discussion.The more frequently users send danmaku,the less time they actually spend watching the video.Integrity-first users are more inclined to discuss course knowledge,rather than express learning feelings.【Innovation/limitation】Innovatively integrates the user's danmaku text content and time information,and realizes a more comprehensive user portrait.Follow-up research can continue to explore the differences and connections of user portraits in videos of different topics.
作者
杨阳
余维杰
YANG Yang;YU Wei-jie(School of Information Management,Sun-Yat-Sen University,Guangzhou 510006,China)
出处
《情报科学》
CSSCI
北大核心
2022年第12期161-169,共9页
Information Science
基金
广东省哲学社会科学规划一般项目(GD22CTS05)
中山大学中央高校基本科研业务费项目(19wkpy149)。
关键词
用户画像
弹幕分析
在线学习
信息行为
主题模型
user portrait
danmaku analysis
online learning
information behavior
probability topic model
