Family Tragedy or Gender-Based Violence: Towards an Automatic Observatory of Femicide Reporting in the Dutch Press

Gosse Minnema, Sara Gemelli, Chiara Zanchi, Tommaso Caselli and Malvina Nissim

Femicide – the killing of a woman because of her gender by a man, usually by an intimate partner – is a major societal problem, but is not always recognized as such. According to the Atria Institute, a femicide is perpetrated in the Netherlands every eight days [1], putting the country in the European top-3, above countries like Italy and Spain where femicide is widely seen as a societal problem and frequently gives rise to large-scale street protests. By contrast, in Dutch media discourse, there seems to have been much less attention on the topic. A Google trends analysis [2] suggests the term “femicide” was little-known in the Netherlands until 2020, when several articles (e.g. [3]) discussing the topic – and explicitly using the term – appeared in the mainstream press. This does not mean that femicides are ignored by the media: reports on murders of women by their (ex-)partner are frequent, especially in regional media. While no quantitative data is available yet, it appears that news coverage often frames these murders as “family dramas” (a), or takes the perspective of the perpetrator by explicitly mentioning his feelings (b), but rarely discusses them in the context of wider societal patterns of violence against women.

(a) Politie bevestigt: man vermoordde zijn vrouw tijdens familiedrama Convivastraat” [“Police confirms: man murdered his wife during family drama in the Convivastraat”],, 2022-03-31
(b) ‘Hij was jaloers.’ Vriend van doodgestoken studente (21) in Leeuwarden zegt dat verdachte uit Leek (27) haar niet los kon laten. [“‘He was jealous’. Friend of female student (21) who was stabbed to death in Leeuwarden says that the suspect from Leek (27) couldn’t let go of her”]. Dagblad van het Noorden, 2022-03-07.

Cognitive and psycholinguistic research into the reporting of femicides and other forms of gender-based violence has shown that different choices of linguistic expressions that foreground or background the agentivity of perpetrators can influence the extent to which readers perceive the perpetrator is held responsible and blamed for the violence, and that agentivity-backgrounding expressions are highly prevalent in media discourse (Bohner, 2001; Meluzzi et al., 2021; Pinelli and Zanchi, 2021). Recently (Minnema et al. 2021, Minnema et al. 2022, Minnema et al., in prep) we have introduced a computational approach to analyzing media discourse about femicides that combines frame semantic parsing and syntactic analysis to identify agentivity-backgrounding descriptions of femicide, and conducted a perceptual study aiming at predicting perceived responsibility from text. However, so far, our work has focused on Italian due to the availability of relevant, high-quality corpora about femicides in Italy. In the present contribution, we will discuss preliminary findings of efforts to semi-automatically create such datasets and transfer computational tools developed for Italian to the Dutch context. Specifically, we will present an online observatory that automatically detects new media reports of femicides in near-real time and provides several types of automatic linguistic and perceptual analysis of these reports. A preliminary version of the tool is available at: