In June 2014, we conducted a pilot study using the tweets of 100 Twitter users who self-identified as teachers. These tweets were collected over eight days using ScraperWiki. Using a sub-set of 26 user feeds that included high volume, mid-range and lower volume tweeters, the PIs analyzed 1971 unique tweets and developed a coding scheme. We further grew our list of self-identified teachers to 507 accounts.
Using Gephi we analyzed the network of the 100 accounts and found that teachers were talking to each other on Twitter. Through our reading of tweets, we found that on Twitter teachers discuss their teaching practice, develop networks of pedagogical and political support, and communicate with non-educators about the work of teaching.
Beginning August 2015, we collected the tweets of 507 self-identified teachers, from which we collected 550,508 tweets, which included 4,335,085 non-stop words. Based on our pilot study, we created a dictionary of words of interest, and a list of moral words. Using these dictionaries, we sorted the accounts using 10,000 trials in Autoclass and came up with five classes. In order to explore the distinction between the classes, we visualized the tweets containing moral words from the entire network and then each individual network in order to explore the connections amongst and across classes. We also mapped high levels of influence of particular words across classes.
Our next phase will target specific groups of tweets for deeper analysis into their content to understand more about the clusters, the content of their tweets, and how teachers are using Twitter.