Emotion recognition for study into populist radical right

InterTextueel is glad to announce its cooperation in a project which will allow us to further improve our sentiment recognition tool for social and political research. Ernst van den Hemel (University of Utrecht) is studying the nature of postsecular society in Europe. He will present a paper in July on the position of religion in populist radical right at the Ian Ramsey Centre for Science and Religion (University of Oxford). He has asked our assistance in helping prepare a dataset and applying automated emotion recognition on it. At the time of writing, we have only taken the first steps to gather data and discuss the basic outline of his research, but we will publish details as the research progresses. By doing so we hope to present an effective demonstration of our tool as well as explore some of its limitations. There are specific challenges with using sentiment analysis on political language, which we shall come to face and discuss.

The dataset we will work on initially consists of tweets from both prominent members of the Dutch populist radical right, and common twitterers who explicitly express their support for it in their online profile (and vocally tweet about it). Some of the leaders of new populist radical right parties or movements self-identify with what they call the “Judeo-Christian tradition”. Van der Hemel’s research will start with questioning whether this positive (but secular) self-identification with the Christianity or Judaism is shared by many of the supporters of the populist radical right.

A far right protest in germany in front of a church.


We’ll proceed by searching in both datasets for mentions of Judaism, Christianity or Islam. This will create several sub-datasets whose emotional “fingerprint” can be compared quantitatively. If there are substantial differences, we will certainly ask why. Qualitatively, we can sort our dataset by the emotions detected in it by our software (for example: display tweets sorted by anger in descending degree) and use the tool to get a quick grasp of the type of content and the emotional extremes we may find in our corpus. Emotion recognition software can be used as a heuristic method to learn which questions may be interesting to ask and/or to help you browse through enormous amounts of data when it isn’t feasible to do it by hand.

We have been discussing the potential of this tool for political and social research with Ernst van den Hemel and we would love to do more broad-based and far-reaching studies in the future, making use of more online (Facebook) and semi-offline sources (parliamentary documentation, for example). We’ll update this research post when we have started to make progress.