基于标签与短评数据的用户标注行为特征分析

Analysis on Users Labeling Behavior Characteristics Based on Tags and Short Comments Data

[目的/意义]基于标签与短评文本数据探究用户的标注行为规律,为优化社会化标注系统的功能和服务提供参考和理论依据。[方法/过程]采用自主编写的Python代码抓取“豆瓣读书”平台健康类目下的图书标签与短评数据,运用统计学方法和自然语言处理技术对用户的标签标注行为与评论行为进行分析。[结果/结论]用户偏向使用字符长度为2~5,词性为名词、动词与形容词的标签对健康类图书信息进行标注,在标签词的选用上呈现出一定的用词规律和词性偏好;无论从单本图书还是整体图书来看,短评文本的情感值与资源的星级评分都具有显著相关;从整体图书而言,星级评分与评论有用性感知得分呈极弱正相关,短评文本的情感值与评论有用性感知得分呈极弱负相关。用户对信息资源的文本评论、星级评分和有用感知行为都具有评价资源的作用,并且它们之间具有一定关联性。本文据此为社会化标注系统平台提高用户体验提出相关建议。

[Objective/Significance]To explore the rules of users’labeling behavior based on the data of tags and short comments,and to provide reference and theoretical basis for optimizing the functions and services of the social labeling system.[Methods/Processes]Through the self-developed Python code this paper captures the book tags and short comments data under the health category of the“Douban Reading”platform,applies statistical methods and natural language processing techniques to analyze the user’s tag labeling behavior and comment behavior.[Results/Conclusions]Users tend to tag health books with characters lengths of lengths 2-5 and part of speech of nouns,verbs and adjectives,Users show certain lexical rules and parts-of speech preference in choosing tag words;Whether from the perspective of a single book or the whole books,the emotional value of the short review text is significantly related to the star rating of the resources.In terms of the overall book,the star rating was extremely weak positive correlation with the review usefulness perception score,and the emotional value of the short review text was extremely weak negative correlation with the review usefulness perception score.Users’text comments,star ratings and useful perceptual behaviors on information resources all play a role in evaluating resources,and they are related to each other.Therefore,this paper puts forward relevant suggestions for improving user experience of social labeling system platform.