Abstract
OntoAndalus aims at constituting a shared conceptualisation of the domain within a future multilingual terminological resource targeted at experts and students of Islamic archaeology. The present version of OntoAndalus is aligned with DOLCE+DnS Ultralite (DUL), an established top-level ontology for the Semantic Web. This article describes the modelling assumptions underlying OntoAndalus, as well as the more relevant design patterns (i.e. artefact types, events and individuals). The latter are exemplified through relevant case studies in the domain, namely those of lighting artefacts, the life cycle of pottery and the several descriptions of Vaso de Tavira.
Introduction
The challenges of developing multilingual terminological resources in the Semantic Web and integrating them in the Linked Open Data Cloud have led to ever closer ties between terminology and ontology engineering [9,27,35]. As of late, computational ontologies have been put forward as building blocks of knowledge-based terminological resources, from healthcare and medical science to cultural heritage and the humanities [4,28]. These resources, which have been referred to in the past as ‘terminological knowledge bases’ [22], are anchored on interdisciplinary work involving domain experts, terminologists/linguists and knowledge engineers. The ties drawn between terminology work and knowledge representation allow, for example, to draft natural language definitions based on formal definitions/descriptions of classes, which could lead to a greater consistency and overall quality of terminological resources. It is therefore paramount – especially in multilingual terminology work – to distinguish language-specific aspects from the conceptual dimension of terminology, which pertains to extra-linguistic domain knowledge [31].
This paper describes work carried out towards the creation of OntoAndalus, an ontology of al-Andalusian pottery artefacts, which is a subdomain of Islamic archaeology [6]. OntoAndalus aims at constituting a shared conceptualisation of the domain, whose elements may be denoted by terms in several languages within a future ontoterminological resource. The creation of this resource is intended to meet the challenges noted by domain specialists with regard to terminological harmonisation (starting with Portuguese and Spanish) and to help further research in archaeology from an ontological point of view [1,2,5]. It is thought that several observations made in this paper will lead to future research in the domain (e.g. artefact functionality, composition, uncertain knowledge in artefact descriptions).
With regard to the development of OntoAndalus, a top-down approach was followed by reusing a foundational ontology, namely DOLCE+DnS Ultralite (DUL). DUL allowed for a rich conceptualisation of the domain, including important topics in archaeology such as artefact typology, events and techniques in the pottery life cycle, as well as the description of individual artefacts. This paper will expand on several examples in each of these topics along with some basic assumptions about the domain.1
The ontology file is available at:
The paper is structured as follows. Section 2 describes the modelling process and explains the assumptions made in OntoAndalus with regard to some of the more important concepts in artefact description (e.g. parts, qualities). Section 3 details the main design patterns which were followed based on several examples. Finally, Section 4 provides a few notes on evaluation.
OntoAndalus was developed from the interpretation of a Portuguese and Spanish corpus of specialised texts in the domain of interest, along with English textbooks and reference works. The more specialised texts consist of conference papers, journal articles, theses and monographs in which the description, typology and terminology of Islamic pottery artefacts is heavily featured as a subject matter. The artefact typologies of the CIGA group [5] and of Rosselló-Bordoy [29,30] were especially important in order to put forward a unified typology of artefacts.
Protégé [23] was used for developing the ontology, while OWL was chosen as a modelling language. This is thought to promote interoperability with Semantic Web and Linked Data technologies, which could, in turn, facilitate the integration of OntoAndalus as the conceptual backbone of a future ontoterminological resource in the domain of interest.
DOLCE+DnS Ultralite (DUL) was selected as the foundational ontology for the development of OntoAndalus. DUL allows for a rich conceptualisation of the domain while also taking into account more practical considerations of ontology development (e.g. interoperability in the Semantic Web). DUL is a streamlined version of DOLCE-Lite, the original translation of DOLCE into OWL. Compared to DOLCE-Lite, DUL features more accessible terminology, a lighter axiomatisation and integrates a constructive Descriptions and Situations module for social and cognitive entities [10]. The latter was especially important for modelling individual artefact descriptions (Section 3.3).
OntoAndalus is developed based on the latest available version of DUL (3.31), which is imported by Protégé through the
Classes are named in upper camel case (e.g.
Object properties are named in lower camel case (e.g.
Individuals are named in lowercase with underscores (e.g.
Labels in Portuguese and Spanish are assigned to the core elements of the ontology, such as the classes representing artefact types. These labels – declared through the
Natural language definitions of classes are also included in Portuguese and Spanish through the
Classes and collections
It is generally accepted that there are several categories of archaeological data to take into consideration in research design [3]. These include some basic concepts in archaeology, such as those of artefact and ecofact. From an ontological point of view, these categories of archaeological data are roles assumed by material objects. Roles are a major category of anti-rigid concepts in ontology design, which is exemplified in the OntoClean method for analysing subsumption hierarchies [16]. Therefore, a distinction should be made between ‘artefact’ in the anti-rigid sense of ‘archaeological artefact’ and ‘artefact’ in the rigid sense of ‘designed artefact’. No object is necessarily an archaeological artefact, only becoming one under certain conditions and for a certain period of its existence, e.g. after being recognised as an evidence of material culture in an archaeological site. On the other hand, it does not seem that an object can stop being an artefact in the rigid sense without ceasing to exist.
Going back to the categories of archaeological data, OntoAndalus, assumes
The concept of artefact in archaeology is broader in some respects than the constructs defined in DUL for artefactual objects, namely the classes
In archaeology, however, fragments of designed artefacts, such as potsherds, are considered to be themselves artefacts. Potsherds can also be the subject of classification schemes based, for example, on their clay paste or decorative technique. It is clear, however, that a potsherd is not something intentionally made or selected and, therefore, does not instantiate either
With regard to the descriptions of the relevant classes of archaeological data, the following is proposed:
Since OntoAndalus is focussed on pottery, it is assumed that archaeological artefacts only include designed objects or their fragments. On the other hand, intentionally selected natural objects would be considered as ecofacts.
OntoAndalus further includes the
The pottery manufacture event is further described in Section 3.2.
While the objects in our domain of discourse are archaeological pottery artefacts, the typology put forward in OntoAndalus is a specialisation of

Artefact collections in OntoAndalus. Solid boxes indicate classes, while dashed boxes indicate individuals.
It is standard practice in our domain of interest to classify artefact types according to their function, e.g. as tableware, transportation or lighting objects [5]. In OntoAndalus, this is achieved through collections, which are modelled as instances of
The use of collections avoids overloading the artefact typology with unnecessary classes, e.g. ‘tableware objects’. It also simplifies the hierarchy considerably by avoiding cases of multiple inheritance, e.g. when a given type may belong to more than one functional class. Another advantage of the approach followed is that it avoids having a collection of artefacts for ‘other uses’, which is present in some artefact classifications. For example, alembics and inkwells were placed within one such class in a well-known terminology and typology of al-Andalusian artefacts [30]. In OntoAndalus, these types could be described as members of the
OntoAndalus further includes a collection of artefact components. The purpose of this collection is to include objects such as the lids and holders of storage jars, which are sometimes included in groups of storage/transportation artefacts along with the vessels themselves [5]. Since the lid of a storage vessel is not used for storage (but instead for closing the vessel), it is therefore more appropriate for it to be modelled as a member of the collection of artefact components.
OntoAndalus makes use of the simple hierarchy of part-whole object properties of DUL. The object property

Class hierarchy of
Parthood further raises the questions of overlap and boundaries. DUL includes the symmetric object property
Still on the topic of parts, OntoAndalus follows the basic distinction between parts defined by fiat and bona fide boundaries [34]. This approach is also followed in the original version of DOLCE [21] and in the CIDOC-CRM ontology [18]. Physical features are parts of objects which are arbitrarily delimited. In contrast, independent physical objects are delimited by bona fide boundaries, such as physical gaps or material discontinuities. Therefore, a lid and a pot are independent objects, since they are separated by a bona fide boundary. However, the lid remains part of the pot as an artefact. On the other hand, the lip and the rim of a jar are fiat objects, since they are determined through fiat boundaries in a bode fide object.
Based on the original hierarchy of features in DOLCE, the
With regard to the specialisation of

Salient vessel parts. Vector illustrations by Mariana Tavares.
Orifice, body and base are usually seen as the primary features of a vessel (Fig. 3) [26]. The body is generally understood to be the part of the vessel between the base and the orifice which includes the area of maximum diameter (or greatest enclosed volume). The orifice, or mouth, may be restricted if its diameter is less than the maximum diameter of the vessel, or unrestricted if its diameter is equal or greater than the maximum diameter of the vessel. The overall shape of a vessel is often described as closed if it has a restricted orifice. Otherwise, the vessel is said to be open. In our domain of interest, this practice is generally adopted by the Portuguese and Spanish archaeologists, who refer to open and closed forms rather than restricted and unrestricted orifices [5,13,30].
Vessel shapes are often more complex, leading to several possible configurations of parts. In closed forms, the area between the point of maximum diameter and the orifice or neck can be referred to as ‘shoulder’ or ‘upper body’, while the area below the point of maximum diameter is the lower body. Another relevant feature of some vessel shapes is the extension of the orifice through a neck. The neck is can be understood as a restriction of the opening of the vessel, extending somewhere after the point of maximum diameter of the body [26].
The final salient feature of a vessel is the base, i.e. the underside of a vessel. However, it is sometimes difficult to distinguish the base from the body in round-shaped vessels, in which the base would correspond to a very small area in the underside of the vessel [26].
Secondary features include appendages, which may be applied to several parts of a vessel (e.g. feet, handles, spouts), as well as the rim and the lip. The latter two features are parts of the mouth: while the rim is the innermost part, the lip corresponds to the outward edge of the mouth [26].
The intended definition of
The functional components of artefacts is a further topic of interest in OntoAndalus. For example, instances of the
The intended definition of

Specialisation of the
Besides structural parts, the
A further issue lies in the relationship between parts of artefacts and non-material spaces bounded by these parts. In ontological literature, holes have been considered as ‘superficial entities’ located in their physical hosts [7]. Following this insight, OntoAndalus distinguishes between the orifice and the mouth of a vessel, in that the former is a part of the vessel and the latter is a non-material space located in the vessel. According to this distinction, the orifice can be understood as the part of the interior space of the vessel bounded by the mouth. In OntoAndalus, spaces are modelled as subclasses of
The mouth is therefore understood as a structural part of a vessel, composed of a lip and rim, having a common boundary with a vessel’s orifice. The following description of
One of the more characteristic aspects of DOLCE lies in the distinction between the qualities of individuals (e.g. colour, shape) and their abstract values, which may change over time (e.g. red, round) [21]. In DUL, qualities can be represented as classes and attributes as individuals instantiating the
While this approach is precise when it comes to describing individual qualities, its application for the definition of concepts is less straightforward. However, DUL allows to link any entity to an abstract region through the
The approach followed in OntoAndalus with regard to qualities can be illustrated with the case of function. It is assumed in OntoAndalus that functions inhere in artefacts similarly to standard examples of qualities, such as colour or shape. This is partly inspired by Basic Formal Ontology, where both functions and qualities are seen as dependent continuants [33].
The

Section of the functional space.
Individuals within this space are used in formal descriptions of artefact types. This approach is based on Rice’s analysis of pottery vessel functionality [26]. The
Special purpose, or non-utilitarian, functions can also be defined in OntoAndalus, such as display (e.g. the function of an elite object with an elaborate decoration) and ceremonial use (e.g. a vessel for ritual ablutions). Since parts of the functional space may overlap, it is possible for a function to be partly included in both utilitarian and non-utilitarian regions. For example, ritual ablutions are a specialisation of washing and, therefore, could also be seen as a processing function.
Artefact types
In this section, the case of lighting artefacts will be described as an example of how artefact types can be modelled in OntoAndalus. Lighting artefacts were chosen as a case study because they include some of the more best-known and representative artefacts of the Islamic period in the Iberian Peninsula.
According to Gómez Martínez [13], archaeologists usually recognise four series (i.e. pottery types) within the lighting objects of the al-Andalus. These are denoted by the following Spanish terms: (i) candil; (ii) policandela (or almenara); (iii) lamparilla; and (iv) fanal (or linterna) (Fig. 6).

Types of lighting artefacts: (i) lamp (candil); (ii) multiple lamp (policandela/almenara); (iii) stationary lamp (lamparilla); and (iv) lantern (fanal/linterna). Vector illustrations by Mariana Tavares.
There is some controversy among scholars regarding the definition of at least two of these pottery series, namely the lamparilla and the almenara. The lamparilla series was first defined by Valdés Fernández in his “Kalifale Lampen” article, published in Madrider Mitteilungen in 1984 [13,37]. Valdés argued that certain pottery vessels found in the Madinat al-Zahra site (near present-day Cordoba, Spain) were used for lighting. This hypothesis was based on the resemblance of the findings with Islamic glass lamps that were placed on support plates. Furthermore, while the candil is a common find in Islamic contexts, it is absent from the Madinat al-Zahra site. The lamparilla would have been part of the domestic lighting system used in the palatial city. While these objects were initially thought to be exclusive to Madinat al-Zahra, similar findings have since occurred elsewhere in the former al-Andalus [17].
Gómez Martínez tentatively proposed the existence of a bifrustum-shaped variant, which is a frequent finding in the Iberian Peninsula (Fig. 6-iii) [12]. In later works, this archaeologist expressed serious doubts with regard to the existence of the lamparilla as a lighting artefact. This was partly due to the recovered artefacts not having the expected signs of usage (i.e. burned marks). However, more importantly, since these objects are opaque (contrary to the glass lamps), they would only be effective in any sort of lighting function when completely filled with oil [13].
The controversy is different when it comes to the almenara series. Rosselló-Bordoy first defined the series as a sort of ‘multiple lamp’ (i.e. with multiple fuel chambers) or as a base for holding several lamps, which could be suspended from the ceiling [30]. While scholars recognise the existence of lamps with multiple fuel chambers in metal, its existence in pottery was merely deduced from potsherds. According to Gómez Martínez [13], the almenara series remains ill-defined: while it is true that some potsherds of lamp fuel chambers were part of a larger whole, its specific configuration remains largely unknown.
Since DOLCE quantifies over possible entities, there is no underlying issue in representing theoretical artefact types in OntoAndalus. The possible types of lighting artefacts were, therefore, introduced in the ontology along with the more established candil and fanal types. This has the advantage of providing more choice for particular groups of experts.
In order to facilitate communication and maintain consistency with DUL, each class is identified in English: (i)
The descriptions of the lighting artefact classes are as follows:
Although the qualified cardinality restriction with a minimum cardinality of 1 is logically equivalent to the existential restriction, it is nevertheless useful in helping to draft definitions in a natural language.
According to the specialists, the candil was made for domestic lighting. Its typical components include a fuel chamber, a spout for holding the wick and a handle. Nevertheless, there are objects within this series that do not have handles, which were designed to be suspended. Also, less typical instances of this type may have more than one spout [8,13,30].
This artefact type was meant for lighting in the outdoors. The fuel contained in its chamber would have been solid, e.g. wax. The typical features of the object are based on isolated findings, and include having a closed shape, globular body and a single handle [5,13].
The scarce information available indicates that this artefact would have had several chambers unified by some sort of structure, like a stand or a base. Furthermore, it was to be suspended or, at the very least, left stationary [13,30].

Types of

Inferred hierarchy of
This hypothetical lighting artefact was either left on a table stand or on a discoidal plate, which would have held several instances [37]. In OntoAndalus, the

Class hierarchy of events in the pottery artefact life cycle.
The division of the
When it comes to events, DUL distinguishes between actions and processes. The former require at least one agent executing a task. Processes, on the other hand, do not strictly involve agency and are considered based on their evolution through time.
By the ‘pottery life cycle’ it is meant the totality of events in which individual pottery artefacts participated in. According to Schiffer [32], the life cycle of an artefact has six stages: (i) procurement of raw materials; (ii) manufacture, through various processing activities; (iii) use, whereby the artefact performs its functions; (iv) maintenance, including repair and replacement of parts; (v) reuse, whereby an artefact is repurposed for a different function; and (vi) deposition, which corresponds to the end of its use-life and integration in the archaeological record. Other relevant events in the life cycle include the distribution of newly manufactured objects to their end-users and the recycling of objects (or fragments thereof) [25].
The above-mentioned classes are modelled as temporal parts of the overarching
A key concept for modelling the life cycle of pottery in OntoAndalus is the
The intended definition in English is the following: ‘action by which clay paste is shaped through the application of at least one shaping technique’.

Shaping techniques in OntoAndalus.
As specified in DUL, all actions imply the participation of an agent executing at least one task. Tasks bridge the gap between the ontology of events and that of social objects, more specifically those subsumed by the
Examples will be provided of how shaping and decorative techniques are represented in OntoAndalus. The former include manual shaping, mould shaping and shaping with a potter’s wheel (also known as ‘throwing’) [5,13] (Fig. 10). Throwing has two variants according to the type of wheel employed (slow or fast). The
The subtype of this technique that uses a fast potter’s wheel has the following description:
Or, in English, ‘shaping technique in which a fast potter’s wheel is used to give shape to the clay paste’.
Figure 11 shows the decorative techniques present in OntoAndalus. For example, the cuerda seca technique has the following definition in English: ‘decorative technique consisting of drawing motifs with an oxide manganese solution and filling the outlined spaces with a glaze substance’.

Decorative techniques in OntoAndalus.
This technique has two variants, depending on whether the glaze is applied throughout the surface of the artefact (complete cuerda seca) or only partially (partial cuerda seca). The formal description of the complete variant is the following:
Both steps of the cuerda seca technique, i.e. drawing motifs and applying glaze, are placed in a sequence through the
OntoAndalus also includes attributes for the instantiation of manufacturing techniques. The inclusion of these attributes facilitates the description of individual artefacts and of classes instantiating a technique. For example, the class of glazed pottery artefacts should contain in its extension all individuals that participated in a decoration event in which a glazing technique was executed. As a result, the artefacts have the quality of being glazed, which depends on a decoration event. Or, following the quality/attribute distinction, these artefacts have the quality of being decorated which, in turn, has the value of ‘glazed’.
OntoAndalus includes object properties for representing these types of attributes, namely
A possible application of OntoAndalus consists on modelling individual artefacts in the domain of interest. The subject of individuals is less prevalent in the literature on terminology. Nevertheless, the representation of so-called ‘individual concepts’ and the compilation of their designations is featured in the ISO standards on terminology [19,24].
An important question in this regard is that of uncertain knowledge in artefact descriptions. A notable case in the domain of interest is the artefact known as Vaso de Tavira (‘Vessel of Tavira’), having taken its name from the Portuguese town where it was discovered in the mid 1990’s. The artefact has been restored from multiple sherds and is today a staple of the Islamic collection of the Municipal Museum of Tavira.
There is considerable doubt among archaeologists regarding the use and purpose of the object. It has been described as a basin for ritual ablutions [11], as a sweet basil flowerpot which was used as a dowry gift [36] and as a miniature of a palatial water pool [20]. Furthermore, the specialists do not seem to agree on the approximate date of the object’s creation: some place it in the 9th century [11], while others place it in the late 11th to early 12th centuries [20,36].

Formal description of Vaso de Tavira.

Relevant elements of a description of Vaso de Tavira.
Objective (or at least consensual) information about the artefact includes its Islamic origin, the site where it was recovered and several physical qualities. The Vaso de Tavira is generally described as an inverted frustum-shaped object with a maximum diameter of 420 mm and maximum height of 360 mm. The manufacture techniques employed include the throwing of the vessel itself and the manual shaping of the figures on the rim. The vessel shows painted and incised decoration. The figures on the rim are salient zoomorphic and anthropomorphic motifs. Another salient feature is the elongated nozzle on the rim, which allowed the vessel to be supplied with water through a system of interior channels leading to the orifices in the figures.
The uncontroversial assertions about the artefact pertain to parts (e.g. rim with figures, elongated nozzle), qualities (e.g. size, shape, decoration) and locations (e.g. its provenance). It is also uncontroversial to state that the artefact is an instance of a pottery artefact and of an archaeological find. Other assertions are dependent on the interpretation of the experts and, therefore, on communication acts. The formal description of Vaso de Tavira is illustrated in Fig. 12.
The three above-mentioned interpretations of the artefact [11,20,36], are modelled as instances of
Figure 13 shows the relevant elements of a description of Vaso de Tavira. Based on this particular description, the artefact is a flowerpot for growing a particular herb. It played the role of dowry gift somewhere in the late 11th to early 12th centuries. Since
The approach shown in this section enables the description of individual artefacts where objective or consensual information is separated from subjective interpretations. The latter are still very much relevant in the domain of interest and, therefore, should not be excluded from the ontology.
As elsewhere in OntoAndalus, this information can serve as the basis for descriptions of ontology elements in a natural language. Descriptions of individual artefacts are, in fact, an important activity in archaeology and cultural heritage. The following is an English description of Vaso de Tavira:
Following, Gómez-Pérez, technical evaluation is distinguished from user evaluation of ontologies [14].4
User evaluation remains an activity for future work, which is dependent on the implementation of OntoAndalus as part of an ontoterminological resource.
The OntoClean method, on the other hand, was used in order to evaluate the subsumption hierarchy based on the so-called ‘metaproperties’ of rigidity, identity, unity and dependence [16]. This essentially avoids abusing the class inclusion relation in ontologies, which typically results from conflating several senses of ‘is a’. Following OntoClean led to the realisation that there are several relevant concepts of artefact in archaeology, most notably those of archaeological artefact (one of the basic categories of archaeological data) and designed artefact. Since the types of artefacts represented in OntoAndalus are rigid classes (e.g.
Another important contribution of the OntoClean method is the distinction between several kinds of concepts based on possible combinations of metaproperties. The importance of the distinction between types and quasi-types (based on whether they supply or inherit identity) became apparent with the introduction of the
This article described OntoAndalus, an ontology of pottery artefacts of the al-Andalus which is being developed by using standard Semantic Web technologies. The main purpose of OntoAndalus is to represent domain knowledge within a terminological resource in the domain. To this end, a strong possibility consists on adapting a model such as Ontolex-Lemon, which allows to model lexicons and their relationships to ontology elements [15]. It is also thought that OntoAndalus will help further research in archaeology by enabling the ontological analysis of important topics in this domain, such as artefact functionality (Section 2.3), composition (Section 2.2) and uncertain knowledge in artefact descriptions (Section 3.3).
The choice of DUL as a foundational ontology enabled a rich conceptualisation of the domain. In this regard, several assumptions underlying the modelling process were reviewed, starting with the basic categories of archaeological data and moving on to the topics of parthood and quality.
The design patterns regarding artefact types, events and individuals were then exemplified based on relevant case studies in the domain. The modelling of consensual and controversial artefact types is shown, as well as the classification of Islamic lamp types based on multiple criteria of analysis. With regard to events, the pottery life cycle is modelled, along with its temporal parts and manufacturing techniques. Finally, the matter of individual artefact descriptions is exemplified through the case of Vaso de Tavira, which further illustrates how several competing interpretations of the artefact can be modelled.
Footnotes
Acknowledgement
Research financed by Portuguese National Funding through the FCT – Fundação para a Ciência e Tecnologia as part of the project Centro de Linguística da Universidade NOVA de Lisboa – UID/LIN/03213/2019.
