Extracting Primary Emotions and Topics from the Al-Hayat Media Centre Magazine Publications, Using Topic Modelling and Lexicon-Based Approaches

Abstract

It is well recognized that ISIS have effectively utilized several channels to disseminate their propaganda. Producing a considerable amount of unstructured data, the Al-Hayat Media Centre had a cornerstone role in the implementation of the propaganda which targeted audiences of the Western world. Although there was a considerable effort by researchers to analyse the ISIS’s online propaganda, there are very few studies which try to relate the above-mentioned material with the emotions associated with it. In this paper, we study the influence of all known issues of the Islamic State magazines, published by the Al-Hayat Media Centre, used to disseminate propaganda, and introduced ideas to audiences outside Iraq and Syria. We perform a series of natural language processing techniques, such as Emotion and Polarity Analysis using a lexicon-based approach and topic modelling with Latent Dirichlet Allocation (LDA). We extract deep insights from the terrorist corpus, including the emotions that are dominant in the text, the combination of emotions (patterns) that the magazine authors attempt to evoke to the reader, the main topics of discussions and the overall ‘feel’ and theming of the propaganda. The combination of lexicon-based emotion analysis with the topic modelling further achieves a reduction in the ‘emotion noise’ that can be observed when applying the lexicon-based approach only, making clear which the dominant emotions are. The valuable findings discussed in this paper contribute to the existing Information Warfare literature as they provide a whole new perspective in the subject of ISIS’s propaganda. The method also aims to open an interdisciplinary discussion, with applications in politics, marketing, etc.