Named-Entity Recognition (NER) plays a vital role in extracting critical information from crime-related documents, which is essential for law enforcement agencies. However, the lack of adequately annotated data on real-world crime scenarios hinders the effectiveness of NER in this domain. To bridge this gap, researchers have introduced a new dataset, CrimeNER, specifically designed for zero- and few-shot NER in the crime domain1. This dataset aims to facilitate the extraction of information about crimes, criminals, and law enforcement agencies involved. The development of CrimeNER is significant, as it can enhance the capabilities of law enforcement agencies to analyze and respond to crime-related data. The use of zero- and few-shot learning techniques enables NER models to learn from limited annotated data, making them more practical for real-world applications. By leveraging CrimeNER, researchers can develop more accurate and efficient NER models that can extract relevant information from crime-related documents, ultimately supporting law enforcement agencies in their efforts to combat crime. The implications of this research extend beyond technology, as it can inform policy and security decisions, and impact workforce dynamics. So what matters to practitioners is that the development of CrimeNER and its application in zero- and few-shot NER can significantly improve the efficiency and effectiveness of crime-related data analysis, enabling law enforcement agencies to respond more effectively to emerging crime trends.