Rule-Based Crime Information Extraction on Indonesian Digital News

Crime Information Extraction is a task to extract some entities in the crime domain. Previous researchers have studied this task using rules to extract some crime entities in the English dataset. However, the rules were not very precise, which made the system has miss-classification. The classification error is due to the inability to resolve the name entities. This study proposed a system that can extract crime-related information in the Indonesian language. Indonesian citizens said they need to know the crime information openly based on the Crime Information Need Survey. The extraction is done by creating rules that combine dependency parsing and Part-Of-Speech tagging. Two main methods will be implementing as Crime Classification using Ontology and Rule-Based Crime Argument Extraction. These methods will find five crime entities: crime type, victim, perpetrator, location, and time. The evaluation was conducted by comparing the system output with the data manual labeling. The evaluation results were 60.70
^%
Fl-Measure, 62,43% precision, and 59,06% recall. The result shows that the proposed method still needed to be fixed in some areas, especially in creating a combination of rules. The system still hard to define the perpetrator entities, victim entities, and location entities