Language Use Similarities in Small Dutch Twitter Communities

摘要

In this paper, we show that there are measurable differences between the language use of different small (Dutch) Twitter communities. We measured distances between n-gram probability distributions in 1,000-word samples of tweets by about 13,000 Twitter user communities ranging from 4 to 51 participants, discussing one of five selected broad topic areas, in which strong content-bearing words were masked. We found that the language use is more similar in users from the same community than in users from different communities. This was to be expected for discussions of the same topic, where the community might focus on specific subtopics, but the effect was also found across-topic.