文本分析的新工具:以主题模型的可视化为例  

A new tool for textual analysis based on the visualization of the topic model

作　　者：王涛[1] WANG Tao(Nanjing University,Nanjing 210023,China)

机构地区：[1]南京大学,南京210023

出　　处：《云南大学学报（社会科学版）》2018年第6期75-79,共5页The Journal of Yunnan University:Social Sciences Edition

基　　金：2017年度江苏省第五期“333高层次人才培养工程”的支持。

摘　　要：主题模型是帮助我们理解文献集的重要手段。但LDA的原版算法需要用统计语言学的知识进行解读,不合适人文学者使用。借助py LDAvis、TOM等软件包进行可视化处理,为主题模型的结果解释提供了便利。从本质上看,任何可视化的方法都需要结合研究项目的具体需求,而不能为了绚丽的可视化结果而毫无目的地使用可视化的方法,否则就有本末倒置的嫌疑。The topic model is an important tool to help us understand the corpus.However,the original algorithm of LDA requires statistical knowledge to interpret the results from humanities scholars.We used several Python packages that included pyLDAvis and TOM for visualization,which facilitated the interpretation of the results of the topic model.Essentially,any visualization method needs to be integrated with the specific needs of the research project,not to aimlessly use the visualization methods only for gorgeous visualization charts,or it may have the suspicion of having the order reversed.

关 键 词：主题模型 可视化 数字人文 

分 类 号：K061[历史地理—历史学]