In simpler terms, the article is about how two people independently labeled a set of tweets and measured the agreement between their labels using a statistical method called Cohen’s kappa. They found that they agreed substantially (0.75) on the first 30 tweets, and after labeling more tweets, their agreement improved slightly to 0.78. The authors think this stability is due to several factors, including having a prior discussion about all the data, the brevity of tweets, and the experience of working together on similar projects in the past. They also provide details on the coding schemata used for each research question.
Computer Science, Software Engineering