🖧 A NETWORK OF ONTOLOGIES FOR EUROPEAN POETRY
The starting point for the development of the ontology is the domain model for European Poetry. This model was created through a process of inverse engineering in order to retrieve the informational needs of the community of practise. These are analysis of patterns, establishment of functional and non-functional requirements.The resulting conceptual model presents a great complexity because of its exhaustiveness and due to the various conceptual domains that were explored. It covers both the bibliographical and technical information of the works, and also metadata derived from literary and prosodic analysis. Moreover, it includes elements that are complementary elements of the texts like images and musical notation.
In order to reduce the complexity of the model, we began by identifying knowledge areas and defining a series of subdomains.
Thus, we identified the following subdomains (see illustration below):
- A central subdomain with the concepts and properties for the representation of the poetic work.
- A subdomain with the necessary entities to deal with the aspects of the transmission of the poetic work.
- A subdomain with the necessary concepts and properties to describe the structural elements of a poetic work.
- A subdomain with the necessary concepts and properties to describe the prosodic elements of a poem.
- A subdomain to incorporate the necessary knowledge for the literary analysis of the work.
- A subdomain dealing with musical concepts and properties related to the work.
- A subdomain that deals with the conceptualization of entities complementary to the works, such as illustrations.
- A generic subdomain with the necessary concepts and properties for the representation of agents, places and related events and roles.
The graph below shows the subdomains.
Illustration 1. Model subdomains
Therefore, the development of an ontological model for European poetry has been carried out through a modular design that is derived, initially, from the subdomains identified in the domain model. Subdomains have a well defined semantics but it is necessary to make refinements to determine which of the subdomains corresponds to a complete and independient ontology and which should be merged with others to create a larger ontology.
The modular design results in several ontologies which are connected in a network of ontologies for European poetry. We tooh into account the following criteria:
- i. The classes, relations and axioms of the ontology are thematically related. In this case, it is a question of guaranteeing that the underlying semantics of each class is related to the area of knowledge.
- ii. Weak coupling, that is, the ontology must present sufficient autonomy which means that it is a self-contained ontological module that preserves the relationships with other ontologies (Oh, Yeom, & Ahn, 2011).
- iii. Strong cohesion, that is, the ontology must have the maximum number of properties between classes to achieve a high degree of cohesion in the ontology, which describes the functionality of the ontology and avoid as much as possible the coupling with other ontologies, i.e. the degree of interdependence between ontologies is small.
To measure the degrees of cohesion, we applied a metric proposed in Oh, Yeom, & Ahn (2011), which measures the cohesion of the modules, i.e. the ontologies, independently of the cohesion of the ontological network. This metric takes into account a factor that ensures the quality of the module. This factor is the degree of internal relations in the module, that is, the degree of cohesion present in the module that undoubtedly influence the coupling with other modules. This also makes it possible to verify the logical consistency between the modules and the complete ontology.
Cohesion in an ontology module has to do with the degree of relationship of the classes in the module. Classes are related when they share properties or have connections with other classes. Therefore, the relations contemplated in this metric can be both hierarchical (the properties of the parent are shared) and non-hierarchical (the classes connect to each other).
Formula 1. Cohesion calculation formula
For this measure, we considered the dependencies established with other modules of the ontology (through import), thar are necessary to complete the defined module.
For every subdomain, we carried out different analysis to obtain:
- Identification of the own classes of the subdomain and identification of possible refinements in the definition of the classes
- Identification of the attributes or data properties of each of the classes
- Identification of the object properties that are resolved within the subdomain that is being modeled
- Definition of cardinalities and obligatory restrictions
- In all cases, we considered the reuse and alignment of ontologies and ontological patterns
The network of ontologies
From this process, we identified the following ontologies as part of the network of ontologies for the domain of European poetry
This ontology covers aspects related to poetic works and their manifestations.
The classes PoeticWork, Redaction and Ensemble have been defined for this purpose.
Since it is the core or central ontology of the network, we incorporated classes that are not specific to poetry but that represent a transversal knowledge. These classes complete the relevant information not only for the classes of the core ontology but also for other ontologies of a more specific domain.
The following entities have therefore been identified:
- Person and Organisation, to model the agents that participate in the poetic work with different roles.
- CreatorRole and Role that model the authorship or creation of the works and elements related to the manifestation and transmission like editors or scribed, for example.
- Place and Event, to represent places of origin and mentioned events (and places).
The core ontology is imported by the rest of ontologies of the network. For this reason, besides containing the mentioned classes, it also provides a set of common properties that have the same semantics in all the classes in which they are defined. In this way it is possible to express semantics in an unambiguous way for properties that from this point of view represent conceptually the same thing.
The data properties related to the dating of the work, its manifestations and its transmission, contain different features to capture the specifics of the expression of the date in this domain. We needed to take into consideration the difficulty of dating works according to established formats. Dates cannot always be defined with accuracy and this entails the need of additional elements to describe the dating issues. Therefore, we created a small ontology of dates to better represent datation issues when dealing with historical data and periods.
This module contains the classes and properties that are necessary to obtain information from literary analysis. The classes in this module are Acrostic, Intertextuality, RhetoricalDevice.
This module covers the classes and properties related to the textual structuring of the manifestations of the works. The classes that form it are: Syllable, Line, Stanza, Word and Punctuation, these last two belong to a LexicalUnit hierarchy.
This module contains the classes and properties that model the required information for the prosodic analysis of a poetic text. As in other modules, we defined a hierarchy of classes that models the patterns of different levels and that are oriented to define the recurrence of the pattern followed by the stanzas, the lines and the poetic work itself. The classes that form this module are LinePattern, StanzaPattern, Work Pattern, Métrical Encoding, Symbol, RhymeMatch.
6. postdata-transmission ontology 🔗 (in progress)
This module covers the classes and properties concerned with the transmission of the works. The classes of this module are PrimarySource, BibliograhicSource, Witness, WitnessCollection, Repository, Facsimile, Reading, Apparatus, Location.
7. postdata-music 🔗 (in progress)
The POSTDATA ontology network also takes into account a feature presented in many poetic works, which is the presence of musical accompaniment. In this ontology, we have not sought a detailed representation of the musical characteristics but those that can enrich the text and play an important role as complementary information. The classes are: Melody, MusicalNotation, Performance
8. postdata-additionalFeatures ontology 🔗 (in progress)
In the manifestations of poetic works, elements that increase the expressiveness of the works or add context appear regularly. The aim of this ontology is to cover these aspects. The classes identified in the model are: Paratext, Illustration and Scene.
olo: http: //purl.org/ontology/olo/core#
Referenced ontology design patterns