Nowadays, Gender-Based Violence (GBV) has undergone a normalization process, whereby violent behaviors, by being justified as normal, have become subtle and dificult to recognize. In NLP, GBV has been investigated within the broad topic of Hate Speech detection, distinguishing between the diferent targets of hateful contents. Considering the pervasiveness of GBV and its media representation in our society, the main goal of our research is to explore people's reactions to femicide events, considered the most brutal expression of GBV. In particular, we collected 932 YouTube comments in response to the news regarding Giulia Cecchettin's femicide and we proposed an annotation task through a fine-grained annotation schema that builds upon Ferrando et al. [1] with some modifications. The qualitative analysis of the annotated comments revealed some diferences from the GBV-Maltesi dataset [ 1], especially regarding misogyny, aggressiveness and responsibility attribution. We tested diferent LLMs, investigating their ability to recognize the presence of aggressiveness and responsibility in both Maltesi and Cecchettin datasets and to indicate their target, using diferent prompts. Warning: This paper contains examples of ofensive content.
perpetration of GBV by presenting it as a normal
component of relationships [
Attribution 4.0 International (CC BY 4.0). 1https://eige.europa.eu/publications-resources/publications/eu-g Building on the assumption that certain sociodemoender-based-violence-survey-key-results graphic characteristics of the victims might have an im2https://commission.europa.eu/strategy-and-policy/policies/justi ce-and-fundamental-rights/gender-equality/gender-based-viole 3https://www.unwomen.org/en/articles/explainer/five-essential-f nce/what-gender-based-violence_en acts-to-know-about-femicide
pact on determining the lesser or greater spread of the
news [15] and its perception, this study aims to compare
two cases of femicide involving victims who difer in key In recent years, the escalation of GBV has made
femicharacteristics that shape public perceptions of the event. cide a topic of daily discussion5, exposing people to
We intend to do so by adopting the same methodology news and contents related to this extreme form of
vipreviously developed by [
November 2023 in Padua, Italy. Although Giulia Cecchet- Regarding the narrative strategies adopted in presenttin and Carol Maltesi share some common features, such ing femicides, it has been noticed that news media typas age, skin color and origin, they difer significantly in ically cover the killing of women as isolated incidents others, such as motherhood and job. From an intersec- rather than as parts of a broader context [18] that stems tional perspective, in which diferent axes of identity such from the pyramid of GBV. This narration can be damas gender, ethnicity, sexuality, class, and ability intersect aging as people exposed to these forms of media may [16] and create diferent degrees of discrimination [ 17], struggle to recognize it as a part of a widespread social these sociodemographic dimensions may be relevant in problem [19], causing the persistence of violence. influencing diferent news perceptions in users and merit Several linguistic studies focused on GBV represenfurther explorations in our study. tation in Italian media, creating corpora [20], and em
In addition, the Cecchettin case has been selected be- phasizing dominant strategies and narrative patterns cause of its significant media resonance (due in part to [21, 22, 23]. In this context, Mandolini [19] conducted the young age of both the perpretrator and the victim), a qualitative discourse analysis focused on journalistic and the widespread public and social engagement it gen- narratives of femicide in newspapers which describe diferated, largely due to the active involvement of Giulia ferent attitudes in the portrayal of femicide. In particular, Checchettin’s family. Furthermore, the dataset has been the author highlights discursive strategies that (directly created to allow a diachronic analysis. In fact, comments or indirectly) blame the victim and implicitly excuse the have been extracted throughout to cover the entire se- perpetrator, referring to gender stereotypes and romanquence of events that preceded the discovery of the body, tic love rhetoric. Moreover, other studies focused on the i.e. the kidnapping of the victim and her disappearance responsibility framing in GBV [24, 25, 26], specifically for a week, elicited strong emotional responses from the identifying lexical choices and syntactic constructions public. that overshadow the agentivity and responsibility of femi
The most significant contributions of this work are cide perpetrators in Italian news [27]. Considering that
detailed below: diferent linguistic choices trigger diferent perceptions
and responsibility attributions [24], Minnema et al. [
5As noted by the national observatory managed by "Non Una Di Meno" association (https://osservatorionazionale.nonunadime no.net/anno/), Italy is consistently afected by GBV, reporting 120 femicides in 2023, 115 in 2024, and 48 until June 2025.
3.1. Data Collection
To our knowledge, Ferrando et al. [
Bearing in mind a conjoined work with the two corpora,
GBV-Cecchettin was labeled following Maltesi’s
annotation scheme (See Appendix A), proposed in Ferrando
et al. [
Considering the great quantity of media content released
regarding this femicide, we mainly selected videos
uploaded by nationally relevant news broadcasters (e.g., La
Repubblica,Rai, Fanpage.it), moved by three main
motivations: avoid subjective interpretations and favor factual
information, take advantage of the broad spectrum of
users that navigate the platform daily, and mimic
GBVMaltesi data collection process [
For each of the 33 videos we collected all first level
comments. For the annotation phase we extracted 1,500
examples from the gathered comments, maintaining the
original balance between time-sections resulting in 195, 1,073
and, 232 respectively for the first, second and third
timephases. The selection has been made using BERTscore
[
the already available intersectional labels. • Aggressiveness: we added institutions, male population, and split social network into VSN and PSN to the already present targets. • Responsibility Attribution: we added institutions, male population, physical and psychological factors, and split social network into VSN and PSN to the already present targets. • Humor: we added the possibility to indicate a target among victim, perpetrator, VSN, PSN, media, rape culture/femicide.
In addition, we decided to include two new dimensions, Topic and Extenuations, the former motivated by the interest to monitor which aspects of the case were more discussed within the comments and the latter introduced to identify examples of the two common tendencies when dealing with GBV, which are the justification of the masculine and the victim blaming [14]. Topic proposed nine selectable options (Victim, Perpetrator, Victim and Perpetrator, Perpetrator and Victim, VSN, PSN, Media/Information, Rape Culture, Femicide and Other) while
Extenuations presented four labels to choose from: Misogyny Intersectionality Aggressiveness Agg. Perpetrator Agg. Victim Agg. Social Network Agg. Perpetrator Social Network Agg. Victim Social Network Agg. Male Population Agg. Media Agg. Institutions Agg. Rape Culture Responsibility Resp. Perpetrator Resp. Victim Resp. Social Network Resp. Perpetrator Social Network Resp. Victim Social Network Resp. Male Population Resp. Media Resp. Institutions Resp. Rape Culture Resp. Psycho-fisical Factor Empathy towards the event/ Support Sup. Perpetrator Sup. Victim Sup. Social Network Sup. Perpetrator Social Network Sup. Victim Social Network Sup. Male Population Sup. Female Population Humor Macabre Context
Yes % No % 9.03% 90.97% 4.63% 95.36% 24% 76% 19.19% 80.81% 1.23% 98.77% 0.88% 99.11% - - - 2.73% 97.27%
- 0.41% 99.59% 32.89% 67.11% 22.09% 77.91% 6.55% 93.45% 2.11% 97.89% - - - 0.99% 99.01%
- 4.06% 95.94%
- 28.25% 71.75% - - - - - - - 3.14% 96.86% 3.27% 96.72% 97.51% 2.49%
4. Corpus Analysis • Victimization of the perpetrator: to be selected when comments highlight external factors that portray the perpetrator as a victim of the circum- In the annotation phase, we involved 10 students from stances. a Master’s Degree course in Linguistics, 7 of whom • Psychologization: to be selected when the perpe- self-identified as women and 3 as men, mostly intertrator is described using terms or attributes that ested in GBV-related matters. Each participant annojustify the killing because of a psychological in- tated 750 comments, with all examples being annotated stability. 5 times each. All people involved participated voluntar• Victim blaming: to be selected when, although ily. Throughout the process, we held meetings with the the perpetrator is held responsible for the killing, annotators to clarify any doubt about the scheme. certain assumptions or claims are presented that We excluded all comments that were annotated as not partially or completely deny the victim’s status classifiable by at least one annotator, ending up with 932 as a victim. comments. All examples were aggregated via majority • Dehumanization of the perpetrator: to be selected voting between annotators.
when the perpetrator’s humanity (or part of it) We report all statistics of the corpus in Table 1. The is denied or when the perpetrator is diminished dimensions with the most positive examples are: Support or ridiculed based on psychological or physical (36.2% of the corpus), Responsibility Attribution (24%), characteristic, particularly those irrelevant to the and Aggressiveness (21.6%). We also report the statistics case. regarding the diferent time-parts in Appendix B. Analyzing the three diferent time-phases, we found that during the first week, Support was mainly directed to the VSN (66.7%), while less attention was given to Giu- particularly influential in shaping user responses, with lia Cecchettin, the victim (36.9%), due to her unknown Maltesi being a single mother and sex worker, and Ceccondition as a missing person. Although Turetta was chettin being a student with no children. In fact, despite also intended as such, the perpetrator was already per- the common brutal nature of both femicides, the corpora ceived accountable for the femicide (50%), sharing the statistics reveal notable diferences in the expression of responsibility with his parents (PSN, 22.2%), blamed for misogyny, empathy, aggressiveness, and the attribution how they educated their son. Turetta’s accountability of responsibility within the comments, proving how these also explains the aggressive manifestations directed to characteristics are very influential in how online users him (61.9%). Once the femicide had been uncovered, the perceive the two femicides.
Support was re-oriented towards the victim (67%, while To be more specific, Misogyny is present in only 1.9% towards VSN it was reduced to 43%). In this second time- of the GBV-Cecchettin entries, compared to 9.03% in GBVsection, the institution was perceived as accountable as Maltesi. These results are understandable considering the the perpetrator (institutions and perpetrator: both 33.1%) two victims’ profiles: while Cecchettin was a young unidue to a detail that emerged from the reconstructions: re- versity student close to graduation and, therefore, harder gardless of a witness reporting Turetta’s aggression, the to blame, Maltesi was a single mother and a sex worker, police did not intervene. Aggressive comments against details often mentioned in the comments. This can also Turetta increased (66.4%) because he was both confirmed be noticed in the Intersectionality label, present in 0.5% as the perpetrator and also because he was found still of the comments in the former and 4.63% in the latter. alive despite theories regarding a possible suicide after Consequently, the empathy expressed towards the events committing the femicide. The media also become a target had also been afected, since we register higher support of users aggression (16.4%), who found the pervasiveness for Cecchettin (36.2%) compared to Maltesi (28.5%). The of the report tactless and disrespectful. Moreover, users lower empathy shown towards Maltesi causes users to found it inappropriate to dedicate such great attention to be more ironic when discussing this case (3.14%), while a single victim or a case of femicide, arguing that many less humor is shown in GBV-Cecchettin (1.3%). GBVother events deserve the same visibility. The third and Cecchettin recorded few instances for Responsibility Atlast time-section presents similar results regarding Sup- tribution and Aggressiveness towards the victims: the port (victim: 75%; VSN: 33.3%) while it shows interesting former accounted for only 1.6% of the corpus while the latoutcomes for Responsibility Attribution and Aggressive- ter was entirely absent. In contrast, GBV-Maltesi revealed ness; in both dimensions, the perpetrator (R: 21.4%; A: a higher degree of Responsibility Attribution directed at 34.8%) had been overlooked in favor of, respectively, in- the victim (6.55%), largely caused by her occupation as stitutions (64.3%) and media (45.7%), indicating the users a sex worker. This aspect led many users to perceive overcoming the specific Cecchettin case to reflect on the her as partially responsible for the violence, therefore, role of the State and the media in the GBV phenomenon. justifying the perpetrator’s aggression. In addition, her
The results recorded for Responsibility Attribution and status as a single mother living apart from her child inAggressiveness, especially in the third time-section, high- tensified both aggressive and victim-blaming narratives light how in the last week the users started questioning within the comments, as shown from the entries (e.g., Ha and discussing the wider problem of GBV, going beyond abbandonato il figlio per darsi al porno, un rifiuto umano the specific Cecchettin case to reflect on the role of the giustamente smaltito. English translation: She abanState and the media in both preventing, punishing and doned her son to turn to porn, a human waste rightly narrating the femicides. As the third time-section regards disposed of.). Finally, these findings support our claim the broadcast of the victim’s burial, aggressiveness to- by illustrating how victims’ sociodemographic traits inwards the media is mostly a condemnation for exploiting lfuence users reactions to femicide news and shape their both Cecchettin’s murder and her relatives’ pain for their perceptions of blame attribution. gain. Considering other targets of Aggressiveness, Cecchettin’s PSN is explicitly attacked when his family takes his 4.1. Divergences and Similarities with side, trying to excuse Turetta for the crime. In particular, we observed more Aggressiveness towards Turetta’s fam
GBV-Maltesi ily, because since the perpetrator was a young student In this Section, we present a comparative analysis of still living in the household his parents are perceived the reactions to the femicides of Maltesi and Cecchettin, as partially involved in the crime. In GBV-Maltesi, the highlighting similarities and divergences. authors did not report any examples of attacks towards
As mentioned above, the selection of the cases was Fontana’s family, probably because he was a 44 years
guided by an intersectional approach, focusing on victims old man, responsible for his own actions. On the
conwho presented diverse sociodemographic traits. Among trary, users wrote aggressive comments against Maltesi’s
those traits, motherhood and profession appear to be parents for not supporting their daughter or looking for
As LLMs can be sensitive to diferent formulation of
prompts [
of the term ‘femicide’ and then P1. • P4, which is structured as following: the definition of femicide, the importance of femicide awareness, the description from LaVentisettesimaOra and then P1. her for several months, e.g., Caspita 3 mesi e nessuno si è insospettito che non rispondeva (English translation: Wow 3 months and no one got suspicious that she didn’t answer).
Among the various diferences, the two corpora also present some similarities: Media were rarely the target of Responsibility attribution, and Aggression toward Rape Culture and Victim also show similar outcomes in both corpora, with a 0.41% in GBV-Maltesi and 0.11% in GBVCecchettin for the former and a 1.23% in GBV-Maltesi and absent in GBV-Cecchettin for the latter.
5.1. Tasks
In this Section, we report the experiments we conducted
to demonstrate applications of the resource in NLP. All
experiments have been carried on both GBV-Cecchettin
and GBV-Maltesi. We used LM-Eval-Harness [
We focused on the categories of Aggressiveness and For an example see Appendix C. We used these four Responsibility Attribution as these dimensions are par- prompt structures for all four tasks by just changing ticularly susceptible to the narrative framing of news the description of the task. The description from LaVencoverage. Thus, automatic analysis of users comments tisettesimaOra varied according to the corpora we used can ofer a deeper understanding of how specific narra- (Maltesi’s description for GBV Maltesi and the same for tives influence public perception. Cecchettin).
First, we explain the experimental setting by describing the tasks, listing all models and prompts used. Then, in 5.3. Data Splits and Few-Shot Section 6, we report and analyze the results obtained from the various models across all tasks.
For Aggbinary and Respbinary, we tested the models on the entirety of GBV-Maltesi and GBV-Cecchettin. Meanwhile, for Aggtarget and Resptarget, we only interrogated models on examples that presented at least one target for the respective dimension. For all tasks, we tested the models in both a zero-shot and few-shot setting (in our case five examples). We did not perform any fine-tuning of the models.
Note that, GBV-Cecchettin and GBV-Maltesi have different lists of possible targets. Thus, we change the target list given inside the prompt depending on the dataset used. 0.28 0.25 0.47 0.42 0.33 0.24 0.52 0.46
Gemma
0.45
0.36
0.51
0.47
Aggbinary
Respbinary
Aggtarget
Resptarget
0
5
0
5
0
5
0
5
Maltesi
Cecchettin
Maltesi
Cecchettin
Maltesi
Cecchettin
Maltesi
Cecchettin
Phi
0.21
0.16
0.43
0.38
0.21
0.19
0.44
0.36
we reported the results for the target recognition tasks, many, for example, the dramatic context of femicide and
Resptarget and Aggtarget. All results reported are the aver- comments often citing the violence committed during
age of the F1-macro obtained by models on all prompts the crime. Also, it could be hypothesized that most of
introduced in Section 5.2. the models taken in examination have gone through a
post-processing phase where they are instructed to not
Overall Considerations In general, as expected, mod- generate aggressive or abusive text [
Also, we observe a sharper increase in performance when switching to the few-shot approach compared to the binary tasks. In fact, the majority of models gain 0.2 in F1-macro score.
Binary Classification Almost always, models in a zero-shot setting perform better in Respbinary compared to Aggbinary. We found it interesting as aggressive, and more generally abusive language, is usually a well studied phenomenon, meanwhile, responsibility detection is rather new. Analyzing the outputs, we found that is caused by the fact that models in the zero-shot setting for Aggbinary were generally biased towards the positive label (e.g., Llama predicting the positive label 90% of the times on average). The factors causing this behavior could be Analysis of Recognition for Specific Targets We analyzed the outputs of Aggtarget and Resptarget to understand
Malt.
Cecc.
Malt.
Cecc.
Malt.
Cecc.
Malt.
Cecc. their performances on each possible target. In addition, e.g., Poverina aveva bisogno aiuto anche lei ascoltate le sue we performed a qualitative analysis, investigating model parole quel delinquente andremmo ammazzato mi dispiace behaviors. (English translation: Poor girl, she needed help too lis
First, we calculated F1-macro score (averaged across ten to her words that criminal should be killed I’m sorry) all prompts) for single targets, casting each target as a sin- and Non ci sono parole per descrivere questo schifoso. Mi gle binary label. As the number of possible combination dispiace tantissimo per lei e per la sua famiglia, pregherò between tasks, Few-shot, subset, models and list of tar- per lei (English translation: There are no words to degets is very large, we decided to only focus on the model scribe this lousy guy. I feel so sorry for her and her family, that had the best overall performance across Aggtarget and I will pray for her). This may be due to the presence of cerResptarget, which is Gemma. tain terms or the recognition of the main referent in the
We reported the results in Table 4. Focusing on comments without an understanding of the overall meanAggtarget, the model shows the best performance for spe- ing of the sentence. In fact, in several cases it seems that cific targets, mostly the Perpetrator, PSN and the Media the feminine form of adjective was suficient to recognize with F1-macro averages reaching 0.88. From a qualitative the victim as target, even though the intention was to perspective, this can be caused by the explicitness of the support her and take her side. For GBV-Maltesi dataset, comments that are aimed towards the perpetrator, with the model recognizes the Responsibility attributed to the users expressing hatred towards him and often invoking victim, specifically in comments that blame her for her a punishment consisting of a life-sentence. For instance, own death, citing her life choices, her status as a mother Uccisa da un miserabile vigliacco . Essere orrendo.Giulia living apart from her child, and her job, e.g., Purtroppo le abbiamo perso tanto. RIP (English translation: Killed scelte di vita sbagliate e le sue abitudini la hanno esposta al by a miserable coward. Horrible person. Giulia, we lost male e a tanti rischi (English translation: Unfortunately, a lot. RIP). her wrong lifestyle choices and habits have exposed her
Meanwhile, Resptarget shows diferent patterns, with to evil and many risks) Notably, the ethical judgment the model not correctly recognizing the Perpetrator re- commonly related to "she was asking for it" does not sponsibility in the zero-shot setting. This can be due to appear in the Cecchettin case. the fact that Responsibility Attribution is more subtle and dificult to identify compared to Aggressiveness. In particular, the attribution of responsibility is not always 7. Conclusion ans Future Works conveyed through explicit and direct expressions, but it is often deduced from the context or the femicide event In this paper, we present the GBV-Cecchettin dataset, itself. which collects people’s reactions to the news of Giulia
Model performs well on the institutions label, as com- Cecchettin’s femicide. We chose the topic because of the
ments explicitly attribute the responsibility to Italy’s lack pervasiveness of Gender-Based Violence in our society.
of severe punishments, even going so far as to invoke We further improved the the fine-grained annotation
physical punishment or death for the perpetrator. schema proposed in Ferrando et al. [
In both tasks, we observe that the model does not per- a new femicide case. The GBV-Cecchettin is composed form well in detecting the crucial victim label. In many by 932 comments, annotated by 10 master students. cases, the reference to the victim’s sphere was enough The annotated corpus shows interesting insights, reto recognize her as the target of aggressiveness, even vealing both similarities and divergences between GBVif the intention was completely diferent. For instance, Maltesi and GBV-Cecchettin. In particular, our analysis focused on the diferent perceptions related to misogyny, psychologically or emotionally, ofering support and help aggressiveness, and the attribution of responsibility, em- if they need it. This approach continued throughout all phasizing the role of victims’ sociodemographic traits in stages of the research. shaping those perceptions. In the experimental phase, we tested several LLMs with four diferent prompts to both GBV-Maltesi and GBV-Cecchettin, to investigate Acknowledgments their ability to detect the presence of aggressiveness and responsibility (binary classification task) and to identify We would like to thank all the students involved in the their target from a fixed list (recognition task). The re- annotation task, who worked with professional attitude sults reveal that the former task is easier than the latter. on a dificult topic like GBV.
Aggressiveness binary detection seems to be a harder The work has been partially supported by “HARMOtask then Responsibility detection given the violent na- NIA” project - M4-C2, I1.3 Partenariati Estesi - Cascade ture of the femicide context. In the target recognition Call - FAIR - CUP C63C22000770006 - PE PE0000013 untask, we found that some targets are easier than others. der the NextGenerationEU programme. For example, aggressiveness towards the perpetrator is easier due to the explicitness of the comments directed References towards him.
Despite its contributions, this study has several limitations that should be considered when interpreting the results. First, we reckon the comments selection procedure, although being motivated (see Section 3.1), can be considered inadequate to capture the users’ perception of the Cecchettin femicide. Second, we acknowledge that involving solely Linguistics Master’s Degree students as annotators might lead to biases in the annotated data.
The last two limitations we identify in this research lay the foundation for future work. Having only investigated data drawn from YouTube and recognizing its limitations, we aim to expand our data source in future work, wanting to gather entries from diferent platforms. Lastly, we are interested in exploring other languages and not limiting ourselves to Italian, adapting the fine-grained annotation schema in a multilingual study to develop a more global perspective on how GBV is perceived.
Considering the power of news media in making a difference for human rights in general and women’s rights in particular [18], we strongly advocate the urgency of focusing on how diferent framing of news can lead to diferent online reactions. Therefore, as future work, we plan to study how specific narratives (e.g., terms used by the media) can directly influence users perception.
The Cecchettin dataset was created in accordance with YouTube’s Terms of Service. Among the 10 people involved in the annotation phase, 8 of them are Italian, one Russian and one US-American, all enrolled in a Italian MA Linguistics course. 9 of them claimed to be interested in GBV-related matters, and 5 had already taken part to GBV-related projects. All the annotators involved in this study participated voluntarily, without any incentives or obligation. From the beginning, we met with them several times to ensure that the topic did not disturb them
Here we report the GBV-Maltesi guidelines that served
as the starting point for the scheme used for
GBVCecchettin. All dimension were used in GBV-Cecchettin,
except for the ones that received changes as noted in 3.2.
• Non classifiable: if the comment cannot be
analysed because it is not written in Italian, because
it consists only of emojis, because it is not
comprehensible or not relevant to the topic (any
comment that was marked as NC from at least 1
annotator was removed from the corpus);
• Empathy: whether, in the comment, there are
expressions of empathy in support of the victim, her
family or the event in general (i.e., condolences);
• Misogyny: whether, in the comment, there is a
presence of discriminatory expression against
women, including blaming, objectifying,
discriminatory and sexist practices used towards them
and their life choices. If misogyny is present, we
asked annotators to indicate its target (group or
individual) based on [
In Table 5 we report the various statistics about the different time parts of GBV-Cecchettin. In Table 6 we report Number of Examples Misogyny Intersectionality Aggressiveness Responsibility Support Humor Macabre Context s s e n e v i s s e r g g A y t i l i b i s n o p s e R t r o p p u S
Perpetrator Victim PSN VSN Male Pop.
Media Inst.
Rape Cult.
Perpetrator Victim PSN VSN Male Pop.
Media Institutions Rape Culture P-F Factor Perpetrator Victim PSN VSN Male Pop.
Female Pop. scription and possible values.
Part 1 189 the various statistics of the targets relative to Aggressiveness, Responsibility and Support for the diferent time parts of GBV-Cecchettin. Note that they can sum up to more than 100% as annotators could select more than one.
In Table 7, we report the prompts for Aggbinary. Other tasks present the layout but slightly change in task de
Dato un commento riguardo un femminicidio, stabilisci se esso contiene aggressività, scegliendo tra: Vero, Falso. [Commento]: [Risposta]: Given a comment about a femicide, determine whether it contains aggressiveness, choosing between: True, False. [Comment]: [Answer]: Leggi il seguente testo: {Descrizione del caso di femminicidio in questione proposta da LaVentisettesimaOra} {p1} Il termine femminicidio viene usato per indicare l’uccisione di una persona di genere femminile nell’ambito di una relazione afettiva o familiare. Il femminicidio costituisce l’atto finale di violenze fisiche e psicologiche ripetute nel tempo. È molto importante essere consapevoli della gravità del fenomeno, soprattutto dal momento che circa ogni due giorni, in Italia, viene uccisa una donna. {p1} Read the following text: {LaVentisettesimaOra’s description of the femicide case in question} {p1} The term femicide is used to refer to the killing of a female gender person in an emotional or familiar relationship. Femicide is the final act of physical and psychological violence repeated over time. It is very important to be aware of the seriousness of the phenomenon, especially because a woman is killed approximately every two days in Italy. {p1} {p3} Il seguente commento riguarda il caso di femminicidio di Nome della vittima. Descrizione del caso: Descrizione del caso di femminicidio in questione proposta da LaVentisettesimaOra. {p1} {p3} The following comment refers to the femicide case of Name of the victim. Case description: {LaVentisettesimaOra’s description of the femicide case in question} {p1} Declaration on Generative AI