Volume 27, Number 9 - 5 September 2022
||This month: September 2022
Disinformation networks: A quali-quantitative investigation of antagonistic Dutch-speaking Telegram channels
This paper presents an empirically-informed exploration of far-right and conspiracist Telegram channels associated with Flanders and the Netherlands. It proposes a typology of antagonistic discourse and narratives that circulate within these public channels. Covering the period between March 2017 and July 2021, this research examines a dataset of 215 public Telegram channels and 371,951 messages, bridging gaps between quantitative and qualitative methods by combining visual network analysis with discourse analysis. This approach reveals an expanding, highly diverse and dynamic network of Telegram channels, marked by overlapping antagonistic narratives, including traces of international conspiracy theories. These observations describe an emerging ‘alt-tech’ platform that harbours and interconnects antagonistic actors and narratives in a specific linguistic and political context.
||Also this month
Toxicity detection sensitive to conversational context
User posts whose perceived toxicity depends on conversational context are rare in current toxicity detection datasets. Toxicity detectors trained on existing datasets tend to disregard context, making the detection of context-sensitive toxicity more difficult. In this paper researchers construct and publicly release a dataset of 10,000 posts with two kinds of toxicity labels: (i) annotators considered each post with the previous one as context; and (ii) annotators had no additional context. A new task is introduced, context sensitivity estimation, which aims to identify posts whose perceived toxicity changes if the context (previous post) is also considered. Machine learning systems were evaluated on this task, demonstrating that classifiers of practical quality can be developed.