This paper presents the results of the research conducted on two Croatian mainstream daily news publications, Jutarnji list and Večernji list, and the ways these publications report on cases of violent crimes in Croatia. Articles from each publication were manually selected and collected in order to build two corresponding corpora – vl_corpus and jl_corpus. Article selection was guided by three predefined criteria: the articles were to have been written between 2019 and 2022, the articles were to describe only those crimes that match definitions of ‘violent crimes’ as those in which the victim is harmed by or threatened with violence (including robbery, sexual assault, assault, and murder), and all selected articles were to contain at least one occurrence of a verb denoting violence. The resulting corpora contained approximately 75,000 tokens and covered 12 major crimes and 72 (jl_corpus) or 115 (vl_corpus) minor ones. A NooJ syntactic grammar model was then developed with the aim of recognizing and tagging verbs denoting violence. Development of the grammar model started with a selection of some general verbs denoting violence (such as silovati [engl. rape], udariti [engl. hit], ubosti [engl. stab], ubiti [engl. kill]) which was then expanded with additional verbs collected from the corpora. Beyond just developing a grammar model that would make it possible to automatically detect and tag verbs denoting violence and describing violent crimes, the aim of this research was to compare the two publications. The comparison found clear differences could between the reporting styles of the two publications, and it was observed that Jutarnji list tended to take a more narrative approach to their reports, relying on a wider range of violence-denoting verbs, whereas Večernji list often used less varied and descriptive verbs, but some of its reports and verb frequencies could be considered attempts at sensationalizing already high-profile cases of violent crime.