In this paper we present the conceptual level of an ontologybased application aimed to support social scientists in their sociological analysis related on crime. Starting from several concrete issues posed by the research team of the Social Observatory on Crime of the University of Sassari, our goal is to build a Semantic Web based tool to collect, organize, and analyze data on crime, as well as for the exploitation of research results by institutions and civil society.
Generally speaking, crime analysis is the activity aimed at nding trends in crimes in order to devise both policies and solutions to crime-related issues. From a sociological point of view, the analysis of crime is addressed to understand how criminal phenomena might in uence social assets in the context of urban and rural areas. Moreover, it aims to identify prevention measures and their potential e ects on social con gurations. In order to do that, data analysis plays a role of paramount importance. Nowadays, information on urban crime is pro tably used by a growing number of local governments to identify major risks and make decisions about the safety of their community.
Computer-assisted qualitative data analysis software (CAQDAS) were
successfully adopted by the sociological research community in order to both
organize and analyze such kind of data { see, e.g., [
In this paper we present the conceptual level of an ontology-based application aimed at supporting social scientists in their sociological analysis related to crime. It represents the rst step towards the development of a tool aimed at organize and manage both quantitative and qualitative data related to this application domain. In particular, the need of such a tool has emerged from several concrete issues posed by the research team of the Social Observatory on Crime (OSC)3 of the University of Sassari. On the one hand, the creation of a SW-based crime information platform might allow stakeholders { institutions and civil society { to easily access to data on crime, for instance by mapping crimes and crime-related issues, and identifying where and how they are occurring, where they are concentrated and why. On the other hand, it can facilitate researchers' work in organizing and managing data collected from di erent sources, such as statistical data and qualitative information from newspapers.
The remainder of the paper is organized as follows. In Section 2 we describe the OSC, while in Section 3 we report the whole process of design and implementation of the presented ontology. We conclude the paper in Section 4 with some nal remarks and discussing future works. 2
The OSC originated in 2012 thanks to an interdisciplinary team (in particular social scientists such as sociologists, psychologists, economists, jurists) which involves researchers from the University of Sassari. It originated from the Urban Study Center (CSU)4, that focuses on urban and territory evolution and transformations; coordinates empirical study and promotes the culture of legality by involving students in actively doing research. The CSU contributes to promoting the adoption of governance approaches by involving private and public bodies as both producers and bene ciaries of research outcomes. It also aims to disseminate activities' results through seminars, conferences, educative courses and scienti c publications. Since 2004, the CSU has started to focus on crime and insecurity in order to observe their impacts on the Sardinian social context and territory. The analysis and monitoring activities are based on data collected from documents provided by justice o cers, newspapers and national statistical reports (e.g., reports of the Italian National Institute of Statistics, ISTAT).
The OSC was built aimed at promoting and generating governance
approaches by trying to involve di erent kinds of stakeholder, such as private
and public bodies. In fact, the inclusion of these actors was supposed to be
relevant both to collect basic information and data, and to de ne concerted
strategies for reducing criminal behaviors and attitudes. For a long time, both
literature and policies have focused on situational crime prevention strategies by
creating \defensible spaces", and less focused on the contrast of the motivations
that encourage deviance. During the last twenty years governance approaches
to crime have been promoted, through collaboration among a multiplicity of
actors (also \external" to the control/protection functions). Following this
approach, the CSU concentrated its e orts on identifying processes for increasing
3 Osservatorio Sociale sulla Criminalita, http://polcoming.uniss.it/node/1133
4 Centro Studi Urbani, http://www.centrostudiurbani.it
degree of key-stakeholders' participation and networking rather than de ning
speci c \interventions" and \architectonic fences". Recently [
The main goal of OSC is to develop an exhaustive database which includes quantitative data (primary and secondary data from di erent sources, e.g., ISTAT and regional prosecutors) and qualitative information obtained, e.g., by analyzing local newspapers (in particular La Nuova Sardegna5 and L'Unione Sarda6). The newspaper consultation has allowed the collection a number of detailed information (e.g., description of places where murders happened, description of authors way of life and their past experiences, connections with other types of crime, etc.) that otherwise would have been di cult to gain and record. However, the consultation refers to those relevant crimes which get newspapers attention such as murders, robberies, attacks, threats, and cannabis cultivation.
A further objective of OSC is to analyze connections between widespread insecurity and crime. In fact, individual and collective behaviors, decision-making and economic activities are often strongly related to the types and the intensity of these connections. Moreover, criminal phenomena should be analyzed in relation to the process of modernization (and its consequences) that has involved the targeted territories by shaping social con guration of urban and rural areas. 3
ocra aims at being the conceptual layer of a Semantic Web based tool focused on the improvement of the processes related to the collection, organization, management, and analysis of data on criminal phenomena in Sardinia by OSC. In the following, we describe design and implementation of ocra (Ontology for CRime Analysis). We can summarize as follows the main steps of this process:
Firstly, we reviewed the semi-structured dataset collected by OSC during a 11 years-long research on criminality in Sardinia. Data on criminal phenomena were collected through speci c forms to be lled with information obtained, e.g., by local newspapers such as La Nuova Sardegna and L'Unione Sarda. The criminal phenomena recorded are murder, attacks, robbery and cultivation of cannabis. Data collection forms were mainly composed of the following information: { Data concerning the newspaper, e.g., name of the newspaper, date, and title of the related article. 5 http://lanuovasardegna.it 6 http://unionesarda.it { Data concerning the crime, e.g., type, place, date, and motive. { Data concerning the authors of the crime and victims, e.g., name, job, age, and records of criminal o enses.
Regarding the second point, we analyzed collected data in order to highlight common terminology, redundancies, and relationships between di erent elements, as suggested by the domain experts of the OSC. The results of this process enabled us to compute a taxonomy { depicted in Figure 1 { related to di erent crimes.
Considering the third point, we proceeded with the choice of the modeling
language analyzing the di erent alternatives o ered by OWL 2. To retain most
of the practical advantages of OWL 2, but to improve on its applicability, in [
Considering the available pro les, we excluded OWL 2 EL because it does not support inverse object properties, while we discarded both OWL 2 QL and OWL 2 RL because they do not support, e.g., existential quanti cation to individuals. Thus, ocra has been developed in OWL 2 DL, and its Tbox is composed of 81 classes, 36 object properties, and 53 data properties. In the following, we describe main classes of the ocra ontology7: ArticoloGiornale (NewspaperArticle) represents the class containing the newspapers information about the speci c crime. Every instance of this class has data properties such as Titolo (Title) and DataArticolo (ArticleDate). Luogo (Place) includes the place where the crime has occurred. Individuals of Luogo are also the places in which the victims and o enders were born or live. 7 The full documentation is available at http://visionlab.uniss.it/OCRA.
Movente (Motive) aims to model the motive of the crime. It has di erent subclasses, each of which is a speci c motive, such as economical, political, revenge, etc.
PersonaFisica (Person) models people related to a speci c crime. It has two sub-classes, namely Vittima (Victim) and Autore (O ender). Every individual belonging to those classes has data properties such as Eta (Age), Sesso (Gender), StatoCivile (MaritalStatus), Precedenti (RecordsOfCriminalOffenses).
Reato (Crime) is one of the central classes of ocra. It has two sub-classes related to the main types of o enses taken into account: crimes involving people or things and crimes related to drug { see below. Reato has di erent data properties such as DataReato(CrimeDate), NumeroVittime(NumberOfVictims), NumeroAutori (NumberOfO enders).
ReatoAPersoneECose (CrimeToPersonsAndThings) In this class are included individuals related to crimes that caused material damage or injure people. ReatoAPersoneECose has two sub-classes: Omicidio(Murder), which includes crimes with homicide, and NonOmicidio(NotMurder). The latter covers a large series of crimes which have not led to murder. NonOmicidio has three sub-classes, each one modeling di erent category o enses, namely Attentato(Threat), Minaccia(Attack), Rapina(Robbery).
ReatoCollegatoAllaDroga (CrimeRelatedToDrug) is the other principal subclass of Reato and is related to all the drug o ences. In particular, we modeled the following two sub-classes of drug o ences: Coltivazione(Plantation) and Detenzione(Possession). Some of the most relevant data properties of ReatoCollegatoAllaDroga are connected to the type and the number of drugs con scated by the authorities, such as Semi (Seeds) and Piante(Plants). Strumenti (Weapons) represents the class containing the instruments used by an o ender to commit the crime. It has various sub-classes, such as ArmiDaFuoco (FireArms), Esplosivi (Explosives) and Veicoli (Vehicles).
Concerning object properties, we brie y describe the ones related to Omicidio, because they enable domain experts to involve in their analysis important data regarding places in which the crime has occurred, o enders, and victims. Noticeable object properties are: { commessoDa(CommittedBy): connects Omicidio to Autore. { haCoinvolto(hasInvolvedIn): relationship between Omicidio and Vittima. { avvenutoA(takesPlaceIn): allows the identi cation of murder's place (Luogo). { commessoCon (hasWeapon): relationship with the murder weapon. { haMovente (hasMotive): it connects the o ense with the motive (Movente).
4
In this paper we described design and development of the ocra ontology, the conceptual level of the ontology-based application aimed to support people of OSC in their sociological analysis related to crime.
Currently, we are developing a data integration layer in order to exploit information coming from relevant external sources, e.g., open data provided by ISTAT and DBpedia. We are also designing a Graphical User Interface to support the ontology population stage, in order to make this process of knowledge acquisition more interactive and dynamic. More, concerning the ontology population, we are studying automated solutions for data collection and insertion.
Finally, we are planning to perform more detailed experimental analysis on the ocra ontologies. Some preliminary experiments have shown us that ocra could be a challenging benchmark for OWL 2 DL reasoners.
Acknowledgments The authors wish to thank the anonymous reviewers for their valuable suggestions, which were helpful in improving the nal version of the paper.