Case Study – The Qatar Blockade
In this chapter, we will look at what happens when we apply one of our classifiers to real data that has not been carefully curated, that we don’t have a Gold Standard for, and that was not the data that we trained the classifier on. This is a real-life situation. You’ve trained a classifier; now, you want to use it. How well do the classifiers that we have looked at so far work in this situation? In this chapter, we will compare the output of a classifier on data collected over an extended period with events in an ongoing news story to see whether changes in the pattern of emotions can be linked to developments in the story and whether it is possible to detect long-term changes in public attitudes as well as immediate responses to key events. This analysis will be divided into three parts:
- We will look at how specific events give rise to short-term changes in the pattern of emotions expressed in tweets
- We will investigate...