Find neighboring words with a GloVe model and R's text2vec package - A analysis of tweets from followers of the Austrian alt-right movement "Die Identiaeren"

The alt-right movement seems to be on the rise, but are they really alternative rights or just plain old Nazis? To get more information on this topic I’ve decided to take a look at their social media footprint. Or to be more precise I want to look at the tweets of people who follow Identitaere_B (the official Twitter account of the Identitäre Bewegung Österreich = Austria’s version of the alt-right movement).

But I don’t want to read all these tweets to find one really shocking tweet which I then exploit as pars pro toto. Mainly, because I really, really, really don’t want to read their tweets, but secondly because those guys are really good at coining such a result just as a bedauerlicher Einzelfall [woeful isolated case]. Instead, we will do what I like best - doing some machine learning. Hence, the result will be replicable and reproducible so that nobody can reasonably claim it fake news.

So, here is the plan:

1. get followers of Identitaere_B
2. get timelines of the followers
3. use the R package text2vec to build a GloVe model
4. use the GloVe model to look up which words in the Identitären-Tweet-Corpus are most similar (the nearest neighbors) to the German words

• Hitler
• Faschismus [fascism]
• Nationalsozialismus [National Socialism]
• Juden [jews]
• Muslime [moslems]