AI methods for assessing crime reporting

Abstract

Online news media play a pivotal role in shaping public perceptions, however, the coverage of events is often selective and subject to editorial or societal biases. This tendency is particularly pronounced in crime reporting, where certain incidents receive extensive attention while others are overlooked, potentially distorting public understanding. This study investigates the representation of crime‐related events in Maltese online news by systematically comparing official police press releases with their corresponding media coverage. To achieve this, a custom system was developed to extract data from four major Maltese news outlets, as well as from official police press releases. The system matches these data sources and assesses their semantic and thematic alignment. The findings of this implemented system indicate that only 40.5% of police press releases were reported by the news outlets, suggesting a notable disparity between official communications and media coverage. This discrepancy may point to the existence of editorial biases or inconsistencies in the timing of publication and the dataset used. Further analysis of the matched articles revealed that, despite extensive paraphrasing, there was a strong semantic alignment, indicating that the core meaning was generally preserved despite variations in language. Additionally, a thematic clustering analysis of the news articles identified three predominant categories: political reporting, traffic‐related incidents, and violent crime. Notably, the analysis revealed a significant intersection between crime‐related content and political discourse, indicating that crime is frequently contextualised within broader political narratives. This overlap suggests that media representations of crime are not solely shaped by the events themselves but are also influenced by prevailing political agendas and framing strategies.