In recent years, scholars have been very interested in the relationship between Dante Alighieri and the gurative arts, leading the way to new research paths that have enriched this topic and made it much more complex and interesting. Currently, this knowledge is disseminated on several books and research papers, making it unmanageable to systematically overview the literary and artistic background of Dante and to gain a sound understanding of how this background was gradually constructed over time. We present OntoComedySources, an ontology for the primary sources of Dante Alighieri's Commedia, both literary and iconographic, with a focus on the latter. We developed our ontology using the Web Ontology Language 2 (OWL 2), according also to several foundational ontologies widely used in the ambit of the digital humanities, such as the CIDOC Conceptual Reference Model (CRM) and the Semantic Publishing and Referencing Ontologies (SPAR Ontologies). Our main goal is to support academics in developing and consulting a digital encyclopedia of La Divina Commedia's primary sources. In addition, we expect that OntoComedySources can easily be adapted and extended to other works and other authors.
The relationship between La Divina Commedia and gurative arts is one of the research topics on which scholars have been very active in recent years. Unfortunately, along with information held in the more acknowledged literary sources, such a relationship remains mainly implicit to contemporary readers. It is the task of specialized scholars to analyze and reveal these hidden connections.
Semantic web o ers well established tools and technologies to deal with such issues. Indeed, it provides formal languages for knowledge representation, such as RDF Schema and the Web Ontology Language, which enable one to connect data and information from di erent sources at a global level, thus enhancing coherence, integration, and dissemination of knowledge. Moreover, dedicated automated reasoning systems permit to verify the consistency of representations, ? Copyright c 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). query datasets, and infer implicit information from what has been already dened. Formal representations of application domains are called ontologies.
Recently, ontologies have been used to represent and organize primary sources
in Dante Alighieri's works [
In this note, we present OntoComedySources,3 an ontology aimed at representing the iconographic and literary sources of Dante Alighieri's Commedia, with particular emphasis on the former ones. Our ontology categorizes the various kinds of sources and connects them with the corresponding text fragments in the Commedia. To this purposes, we make use of di erent standard foundational ontologies in the digital humanities such as the CIDOC Conceptual Reference Model (CRM), the Open Annotation Core Data Model (OA), the Dublin Core ontology (DC), and some modules from the Semantic Publishing and Referencing (SPAR) Ontologies.
We chose CIDOC to model knowledge related to iconographic works, also in
consideration of the fact that it was successfully used in the de nition of
ontologies devoted to iconography [
The primary goal of our e orts is to provide the means necessary to support academics in developing and consulting a digital encyclopedia of La Divina Commedia's primary sources. Digital representation of information sources can be a powerful instrument in the understanding of classical literature. In the case of the Commedia, it would provide signi cant bene ts in the di usion, use, and comprehension of knowledge hidden in several publications, archives, and commentaries fragmented across the globe and the web.
With little e ort, our ontology could be readapted and extended to other literary works and other classical authors. 2 2.1
Semantic web and ontologies Semantic web, conceived by Tim Berners-Lee and presented in 2001, is an evolution of the web in which machine-readable data can be queried and manipulated by software agents on behalf of human agents. In such a vision, web information is explicitly interconnected and equipped with meaning, so that it can be automatically processed by machines, and data can be accessed, integrated, and modi ed at a global level, resulting in increased coherence and dissemination of knowledge. In addition, by means of automated reasoning procedures, it is possible to extract and process implicit information present in data, thus permitting to gain a deeper knowledge of the domain. 3 https://github.com/AndreaDeDomenico97/OntoComedySources
The Word Wide Web Consortium (W3C) recommends OWL, a family of
knowledge representation languages based on Description Logics (DLs) [
OWL, currently in its version 2.1, is grounded on triples or statements, each one representing an atomic unit. These are ways to connect either two entities, or an entity and a data-value, each one represented by an Internationalized Resource Identi er (IRI), i.e., a sequence of characters that unambiguously identi es a resource within a speci c context.
Entities represent the primitive terms of an ontology and are identi ed in a unique way. They are individuals (actors), properties (actions), and classes (sets of actors with common features). Properties are of two types: object-properties and datatype-properties. Object-properties relate pairs of individuals, whereas datatype-properties relate individuals with some data type values, such as strings and numbers.
OWL expressivity can be extended by adding Horn-like rules by means of the Semantic Web Rule Language (SWRL).4 Rules have the form of an implication between an antecedent (body) and consequent (head), intending that whenever the conditions speci ed in the antecedent hold, the conditions speci ed in the consequent must hold as well. 2.2
Foundational ontologies used in OntoComedySources In order to implement a good modeling approach and facilitate future integrations with other ontologies, we constructed OntoComedySources according to the following foundational ontologies, widely used in the digital humanities. { The CIDOC Conceptual Reference Model (CRM) is the international standard for the controlled exchange of cultural heritage information since 2006. It provides a general speci cation which can be adopted in any cultural heritage context to construct a semantic web-based information system, to serve as a guide for good practices of conceptual modelling, and to improve information sharing. There are several institutions that successfully implement CIDOC such as galleries, libraries, museums, archives, as well as any other cultural environment based on cultural heritage data that publishes 4 https://www.w3.org/Submission/SWRL/ and shares its information in the semantic web formats. The CIDOC core covers several general aspects of cultural information, such as material and immaterial entities, events, space, and time. Such general concepts can effectively represent iconographic works. { The Dublin Core (DC) ontology is part of the DCMI project (Dublin Core Metadata Initiative), now an ISO standard. It is composed by a core of terms essential to the description of any digital material, as well as physical resources such as artworks or books. We used it to de ne properties for designating the creator of a literary or iconographic work and the author of a commentary. { The Semantic Publishing and Referencing (SPAR) is a family of ontologies for the description of any aspect of bibliographic resources. Speci cally, we used the following SPAR ontologies:
FRBR (Functional Requirements for Bibliographic Records model ): a model developed for the description of essential aspects of bibliographic records; FaBiO (FRBR-aligned Bibliographic Ontology ): an ontology describing entities that are published or potentially publishable (e.g., journal articles, conference papers, books) and that contain or are referred to by bibliographic references; C40 (Citation Counting and Context Characterisation Ontology ): an ontology that permits to record in-text citations of a cited source, together with their textual citation contexts; DoCO (Document Components Ontology ): an ontology that provides a structured vocabulary written of document components, both structural (e.g., paragraph, section, chapter) and rhetorical (e.g., introduction, discussion, acknowledgements); eFRBRoo: an OWL-DL implementation of an object oriented version of FRBR called FRBRoo. { The Open Annotation Core Data Model (OA) speci es a framework for modeling associations between related resources or annotations. It is helpful for representing commentaries to literary works and for explicating relationships between them and the resources they refer to. 3
The ontology OntoComedySources As illustrated in Fig. 1, the structure of the Commedia is modeled by the classes Verse, Canto, and Cantica. These are subclasses of doco:Line, doco:Chapter, and fabio:Book, respectively. Naturally enough, a verse is contained in a canto and, in its turn, a canto is a part of a cantica. This is expressed by the object properties isVerseOf, isCantoOf, and inCanticum, along with their inverses shown in Fig. 1. The properties just mentioned are all subproperties of the two FRBR properties frbr:hasPart and frbr:isPartOf.
In the following diagrams, classes are represented with oval-shaped borders, while primitive data-types are delimited by rectangular boxes. Object properties and data properties are denoted with solid lines, whereas dotted lines designate subclass relationships.
The name of a class or property drawn from an existing foundational ontology is written in the form hpre xi : hnamei, where the pre x, which is omitted when referring to new entities, stands for the namespace of the ontology.
subClassOf Verse doco:Line inCanticum isCanticumOf
Canto
doco:Chapter isVerseOf hasVerse
The class Fragment, subclass of efrbroo:ExpressionFragment, occupies a central role in the ontology, as each of its instances refers to a speci c textual fragment of La Divina Commedia. Its content is expressed by the data property bodyChars.
Based on the class oa:Selector from the Open Annotation Core Data Model, we are able to make explicit the exact position of the corresponding fragment in the text (see Fig. 2). Speci cally, if a certain fragment is located in the interval from the i-th to the j-th character, we can use the properties oa:start and oa:end to model it. Since the speci c type of selector depends on the format of the information which one is working with, we used the subclass oa:TextPositionSelector.
Every fragment has a starting verse and an ending verse; furthermore, in principle, a fragment could refer to smaller fragments within it. Such knowledge is represented by means of the properties startVerse, endVerse, and composedOf. Notice that both the rst two properties are subproperties of oa:hasSource.
To explain the connection between a text fragment and the literary or iconographic source it refers to, we used the class oa:Annotation. As shown in Fig. 3, every instance of the class oa:Annotation is paired with an instance of the class oa:TextualBody via the property oa:hasBody. A textual body has a date and a textual content. Every annotation is written by a specialized scholar and in our ontology we specify this fact with the property dc:contributor, whose range is the class Scholar, subclass of foaf:Person. The classes oa:Annotation and efrbroo: Expression
Fragment oa:TextualBody are related to the class Fragment via the properties oa:hasTarget and hasCitingFragment, respectively.
The association between a fragment and the work cited in it is modeled by the property c4o:cites, belonging to the Citation Counting and Context Characterisation Ontology (C4O). The range of the property c4o:cites is the class efrbroo:Work. In its turn, such a class is paired with the class Author through the property dc:creator and with the class skos:Concept via the property dc:subject. The instances of the class efrbroo:Work are the literary and iconographic sources we are interested in, while the class skos:Concept is used to specify the thematic area of the related work (e.g., Patristics). The meaning of the other classes is self-evident. Fig. 3. How fragments, annotations, and cited sources are linked with each other. hasDate 15/10/2016 hasCommentary hasCitingFragment
hasContent "cfr. Matth. 7, 13: «spatiosa via est, quae ducit ad perditionem»" dc:contributor
subClassOf oa:Annotation
Scholar oa:hasBody
oa:hasTarget oa:TextualBody c4o:cites
dc:creator Fragment
As depicted in Fig. 4, the class efrbroo:Work is a superclass of both classes LiteraryWork and IconographicWork. Besides, LiteraryWork contains the class Manuscript, while IconographicWork is a superclass of the four classes Fresco, Mosaic, Miniature, and PanelPainting. The properties citeAsLiterarySource and citeAsIconographicSource are both subproperties of c4o:cites. The former has LiteraryWork as range, whereas the range of the latter is IconographicWork.
To specify the spatial dimensions of an iconographic work, we make use of the class Length, subclass of cidoc:Dimension, along with the properties hasDimX and hasDimY. Each instance of Length is associated with an instance of the class cidoc:MeasurementUnit via the property cidoc:hasUnit.
To model the datation of an iconographic work, we utilized the CIDOC class cidoc:TimeSpan and the CIDOC property cidoc:hasTimeSpan, as well as the properties startDate and endDate, both subproperties of cidoc:ongoingThroughout. To register where the works are currently preserved, we employed the classes Site and Region, subclasses of cidoc:Place. The properties inSite and inRegion, whose meaning is obvious, are subproperties of cidoc:hasCurrentPermanentLocation.
A miniature is a small illustration contained in an ancient manuscript. In OntoComedySources, this is modeled by the property isContainedIn. Since we often have no knowledge concerning the author of a manuscript, we included the class Antiquary and the property acquiredBy to designate the antiquarian responsible for the discovery and/or the preservation of a given manuscript. hasLocation inRegion
hasDimX inSite Site Region cidoc:hasTimeSpan
cidoc:
TimeSpan startDate 1304
endDate 1306 As illustrated in Fig. 5, the ontology provides us with the entities to describe the materials employed in the execution of the mosaics and paintings, along with the techniques used. The properties typeOfTesserae, typeOfSubstrateMosaic and typeOfSubstratePainting connect the classes Mosaic and PanelPainting with the class Material. All of the three properties have cidoc:consistOf as a superproperty.
In our ontology, the class cidoc:DesignOrProcedure is a superclass of the class Technique which, in its turn, has as subclasses the classes PaintingTechnique and MosaicTechnique; these are linked with the classes PanelPainting and Mosaic by the properties usedTechniquePainting and usedTechiqueMosaic, respectively.
In order to di erentiate among the di erent types of mosaics, we exploit the class MosaicType, subclass of cidoc:Type, and the property typeOfMosaic, subproperty of cidoc:hasType. typeOfTesserae
Mosaic typeOfMosaic
MosaicType Material typeOfSubstrateMosaic
usedTechniqueMosaic typeOfSubstratePainting
PanelPainting usedTechniquePainting
MosaicTechnique
OntoComedySources contains 46 classes, 76 properties, and 209 individuals. It also includes some SWRL rules permitting to draw simple inferences concerning geographical places in relation with iconographic sources, fragments of the Commedia with their corresponding commentaries, the position of a verse inside the Commedia, and the authors cited in a given canto. Some of the rules are illustrated in Fig. 6. In particular, hasIconographicSource(a; b) means that a canto a contains b as an iconographic source, hasIconographicSourceCitedIn(a; b) holds if a region a contains an iconographic source cited in a canto b, and nally citedIn(a; b) means that a canto b contains a work from an author a as a primary source.
Fig. 7 shows an example of an inference drawn from the third rule in Fig. 6 stating that if canto ?a has as iconographic source work ?b and work ?b is housed in region ?c, then region ?c contains an iconographic source cited in canto ?a. Instantiating the variables ?a, ?b, and ?c with the individuals Pu 9 , standing for Purgatory's canto IX, IWorkMo1 , representing the mosaic depicted in Fig. 8, and Ravenna, respectively, we infer that the town of Ravenna stows an iconographic source referenced in Purgatory's canto IX.
The ontology has been classi ed using the DL reasoners Hermit [
Iconographic sources are harder to nd in mainstream commentaries on the
Commedia. We extracted our sources from a research paper and a monography
written by Alessandro Benucci [
An example of an iconographic source is displayed in Fig. 8. Speci cally, Fig. 8 shows a byzantine mosaic panel, executed in 520-547 A.D. and currently conserved in the Basilica of San Vitale, which depicts Empress Theodora. The corresponding fragment is from Purgatory's canto IX. The scholar suggests that, when writing about the aurora, Dante was inspired by the ornaments in Theodora's crown. In this note we presented OntoComedySources, an OWL 2 ontology for the primary sources of Dante Alighieri's Commedia. We modeled OntoComedySources according to the standard CIDOC CRM ontology in order to accurately represent iconographic sources, and we used the foundational ontologies DC, SPAR, and OA to describe literary sources.
It is widely recognized among scholars that Dante's Commedia was in
uenced by the iconographic art of its time. Indeed, in medieval times, gurative
arts played a signi cant role in educating the population, as it was mostly
illiterate and did not have access to literary texts. Accordingly, we plan to expand the
ontology in order to represent iconographic sources even more broadly, following
some recent works on ontologies devoted to iconography [
We gratefully acknowledge partial support from project \STORAGE|Universita degli Studi di Catania, Piano della Ricerca 2020/2022, Linea di intervento 2".