Automated Statistics Extraction of Public Security Events Reported Through Microtexts on Social Networks

Abstract

Lately, Rio de Janeiro State has been characterized by the occurrence of successive public security events (shootings, assaults, robberies, etc.), causing great insecurity, affecting the daily lives of the population, and worrying public security agencies in the fight against crime. Although the indicators of public security events recently decreased, there is still a feeling of insecurity, while the population uses social networks to notify illegal acts that occurred in their vicinity. Although this collaboration is limited to the crimes that occurred, many published messages are difficult to interpret. Knowledge Discovery is a process of extracting data in an implicit, previously unknown, and useful way that can be applied for different purposes. In this context, Natural Language Processing is a powerful tool that allows the extraction of information from these unstructured data. This work proposes a methodology for automatic knowledge extraction, in the form of statistics related to public security events posted on social networks, particularly the ones occurred in Rio de Janeiro. The main contribution of this work is the proposal of a methodology for the construction of an Information System that allows the collection of statistics of notified public security events. In addition to this methodology, which can also be used in the construction of other Information Systems, this work contributes with a public security event recognition model that has a performance of 95%, and an available dataset that can be used to support other researches, such as: the identification of new behavior patterns, the discovery of hidden knowledge, among other fronts.