On the 12th of December, I gave a lecture on the use of machine learning methods for text analysis at Waseda University. As an example, we used Structural Topic Model to analyze Facebook posts written by Donald Trump and Hillary Clinton during 2016. Below you will find an abstract for the event, slides and a script that will help you reproduce the examples during the lecture
The workshop will present the core principals behind the unsupervised machine learning approach towards text with an emphasis on Structural Topic Modeling (STM). Furthermore, the event will offer practical hands-on experience with topic model analysis of text in R.
Social scientists have used quantitative content analysis for decades to understand the breadth of frames, topics and discourses. The approach can be used for descriptive purposes, such as the study of policy change throughout time. It can also be used for causal inquiries, such the study of the media’s effect on political opinions and attitudes among citizens. Manual content analysis, however, can be both highly time consuming and limited to only a small proportion of the text corpus. The workshop will introduce the recent advancements in Topic Modeling as an attempt to automate content analysis in order enable a study of large data sets. Furthermore, the workshop will lead to a discussion of the benefits of topic modeling in relation to other methodologies within both computer science and social science.
You can find the slides here, and the R script here. You can explore the the data using stmBrowser by opening this link.
.