Historical documents are a relevant part of cultural heritage. It is well-established that this domain is very complex: data are often heterogeneous, semantically rich, and highly interlinked. For this reason, searching and linking them with related contents represents a challenging task. The use of Semantic Web technologies can provide innovative methods for more e ective search and retrieval operations. In this paper we present stole, a reference ontology that provides a vocabulary of terms and relations that model the history of Italian public administration' domain. stole can be considered a rst step towards the development of a knowledge exchange standard on this speci c domain.
Historical texts and documents are considered an important component of cultural heritage. Among other sources, historical newspapers and journals archived in libraries represent a valuable source of information for historians.
In the last years, several portals and digital libraries in the cultural heritage
eld have been enhanced with Semantic Web (SW) technologies in order to
implement methods for more e ective search and retrieval operations. In fact, they
can o er e ective solutions about design and implementation of user-friendly
ways to access and query content and meta-data { see [
One of the main elements of the SW infrastructure are ontologies. They provide a shared and clear representation for a speci c domain and they may play a major role in supporting knowledge extraction, knowledge discovery, and data integration processes. An ontology is usually de ned as a formal speci cation of domain knowledge conceptualization. A Knowledge Base (KB) is composed of two elements, i.e., a Terminological Box (TBox) and an Assertional Box (ABox). The TBox represents the intensional knowledge of the KB and it is made up of classes of data and relations among them. In the following, we use the terms TBox and ontology interchangeably to denote the conceptual part of a KB. The ABox contains extensional knowledge, which is speci c to the individuals of the domain.
In this paper we present stole1, a reference ontology that provides a vocabulary of terms and relations with which it is possible to model the history of Italian public administration' domain. stole represents a rst step towards the development of a knowledge exchange standard on this speci c domain. To the best of our knowledge, this is the rst time that ontological modeling has been undertaken for this speci c domain.
The paper is organized as follows: in Section 2 we describe the speci c case
study that motivated this research. In Section 3 we present the stole ontology
and its design process. This ontology is built on an extended version of stole [
In the last years the way of doing historical research has deeply changed. From traditional research in archives and libraries, which necessarily required the presence on site and had precise limits and constraints, the digitization of historical documents allows new perspectives and ways to conduct research. The use of SW technologies has upset times, costs, methods of historical research, and is also changing the so-called culture of the document.
The heritage of the history of public administration represents a fundamental element to understand the history of the Italian institutions as well as the history of the country in general. One of the main sources used in this eld of research is represented by historical text and documents, including journals and newspapers of the age. Recently, several web sites concerning speci c domain databases on Italian institutions, e.g., the Bibliography of Italian Parliament 2 and Internet Archive 3, o er to the scholars the opportunity to easily access the sources, which are sometimes totally unknown.
The development of the stole ontology responds to the needs of some researchers of Department of History of the University of Sassari which, since the 1980s, have been involved in a project designed to collect and digitalize historical journals regarding the origin and the evolution of institutions, customs, usages, and rules in the Italian public administration. As a result, the arap4 digital archive of the University of Sassari was created and it actually collects a large amount of information about some of the most relevant journal articles published between 1848 and 1946 concerning the legislative history of public administration in Italy. The main goal of arap was to o er to the scienti c community 1 stole is the acronym for the Italian \STOria LEgislativa della pubblica amministrazione italiana", that means \Legislative History of Italian Public Administration". 2 http://bpr.camera.it 3 https://archive.org 4 arap is the acronym for the Italian \Archivio di Riviste sull'Amministrazione Pubblica", that means \Archive of Journals on Italian Public Administration". interested in those documents, such as historians, lawyers, political scientists, and sociologists, a repository of important sources, otherwise hardly accessible.
The journal articles included in arap have a remarkably value for the wealth of information they contain. In fact, starting the research from the journals could make a positive and signi cant contribution to the current knowledge. For example, through the study of these documents it is possible to establish connections between authors, institutions, persons and historical events mentioned in an article. Speci cally, the link between an author and the people cited can reveal a lot, such as the political and cultural reference of the author. Clearly, the relationship between an institution and an historical event can o er support to understand the evolution of public administration. The use of certain concepts or the recurring of names in relation to some historical events can be an indicator of a trend.
In this context, stole represent the core element of the arap digital archive and its main goal is to to model historical concepts and gain insights into this speci c eld in order to support historians in their research tasks. 3
The main steps of the stole ontology design process concerned the identi cation both of the key concepts for this speci c domain, and the proper language for the TBox implementation. Moreover, we populated the ontology, i.e., lling the ABox with semantic annotations. A team of domain experts was involved during the whole ontology development process. In particular, they contributed at the early stage in order to de ne the key issues related to the application domain.
In the rst phase, we detected the main categories of data expressed in the considered historical documents. The results of this process enabled us to detect three categories of elements: 1) Data concerning the authors of the articles, e.g., name, surname and biography; 2) Data concerning the journal and the article, e.g., article title, journal name, date and topics raised in the article; 3) Data concerning some relevant facts and people cited in the article, e.g, people, historical events, institutions. In this speci c domain, historical analysis is based on these categories of information and focused on the interrelations among these data. For instance, the relation between an author and the people mentioned in an article could provide valuable information to historians, e.g., if an author has often referred to Giuseppe Mazzini then it could easily be interpreted that this author was favourable to the republic.
During the second phase, the TBox of the stole ontology has been designed building on some existing standards and meta-data vocabularies, such as Bibliographic Ontology (BIBO) 5, Bio Vocabulary (BIO) 6, Dublin Core (DC) 7, Friend 5 http://bibliontology.com 6 http://vocab.org/bio/0.1/.html 7 http://dublincore.org of A Friend (FOAF) 8, and Ontology of the Chamber of Deputies (OCD) 9. In details, BIBO has been used to describe information about documents, for instance bibo:volume, bibo:issue, bibo:pageStart and bibo:pageEnd, denoting values of volume, issue, page start and page end of an article, respectively. Both BIO and FOAF have been used in order to describe information about people. In the case of FOAF, we reused terms such as foaf:person, foaf:firstName, foaf:surname, foaf:gender. Other information about people, their relationships and the events in their lives have been described using BIO concepts, such as bio:birth, bio:death, bio:event, bio:place, bio:biography. From DC we reused concepts related to the structure and characteristics of a document, e.g, dc:title, dc:isPartOf, dc:publisher, dc:description. Considering OCD, we noticed that it is a relevant source for stole ontology since, for example, most part of the authors of articles related to the history of the public administration topics during their lives were also involved in government activities. In particular, OCD provided us valuable concepts in order to give detailed information about people involved in political o ces. Finally, the usage of these core ontologies allows both extensibility and interoperability of stole with other resources and applications.
In dealing with the modeling language, we decide to use OWL2 DL [
In the following, we describe main classes of the stole ontology. Article represents our library, namely the collection of historical journal articles. Every instance of this class has data properties such as articleTitle, articleDate, pageStart, and pageEnd.
Institution is used to represent the public institutions cited in the articles.
This class contains four subclasses, namely Central, Local, PoliticalInstitution and EconomicInstitution. Notice that Central and Local are disjoint classes. The same holds for PoliticalInstitution and EconomicInstitution. In this way, for example, an institution can be both a local institution and a political institution at the same time. Jurisprudence is a subclass of Article and contains a series of verdicts which are entirely written in the articles. Every individual of this subclass has the following data properties: verdictDate, verdictTitle, and byCourt. Law is also a subclass of Article, and it contains a set of principles, rules, and regulations set up by a government or other authority which are entirely written in the articles. This subclass has data properties such as lawDate and lawTitle.
Event denotes relevant events. It contains ve subclasses modeling di erent kinds of events: Birth and Death are subclasses related to a person's life; 8 http://www.foaf-project.org 9 http://data.camera.it/data
BeginPublication and EndPublication represent the publication period of a journal; HistoricalEvent contains the most relevant events that have marked the Italian history.
Journal denotes the collection of historical journals. This class has data properties such as journalArticle, publisher, and issn.
Person is the class representing people involved in the Italian legislative and public administration history. This class contains one subclass, Author, that includes the contributors of the articles. Every instance of this class has some data properties as firstName, surname, and biography.
Place represents cities and countries related to people and events. Subject is a class representing topics tackled in the historical journals.
In conclusion, there are other examples of ontologies designed for the cultural
heritage domain, e.g., CIDOC CRM [
Currently, we are extending and improving in many ways the implementation
of this work. First, we are dealing with a key issue for historians, namely how
to disambiguate individuals and how to manage changing names, e.g., di erent
people with the same name or institutions that changed name retaining the
same functions. This point still represents an open challenge in this application
domain{ see [
Another open problem relates the ontology population process. For the
current version, stole has been populated leveraging a set of annotated historical
documents comprised into the arap archive. Semantic annotations were
provided by a team of domain experts and individuals were added to the ontology
by means of a JAVA program built on top of the OWL APIs [
Acknowledgments I would to thank the anonymous reviewers for their valuable suggestions, which were helpful in improving the nal version of the paper. Moreover, I would also like to thank my supervisors, Giovanni Adorni and Luca Pulina, for their support, and Salvatore Mura and Prof. Francesco Soddu for the valuable discussions about the application domain.