This paper describes the process of mapping, translation and publication in SKOS format of the RA Thesaurus, a terminological tool developed by the Italian Ministry of Cultural Heritage (MiBACT) as a part of the o cial documentation used for the recording of archaeological nds. In particular, the RA Thesaurus is intended to provide uni ed and meaningful terminology for the description of archaeological objects according to the MiBACT o cial cataloguing standards. After describing the thesaurus, the logic with which it was developed and its internal structure, we report the various phases of the conversion, both from a theoretical and implementation point of view, and the various technologies used for the publication of the thesaurus on the web. This work is a collaborative e ort between PIN and MiBACT carried out under the ARIADNE project.
ARIADNE is a European project focusing on integration of existing
archaeological research data infrastructures to enable the use of distributed datasets and
services by means of new and powerful technologies as an integral component
of the archaeological research methodology. Among other activities, ARIADNE
is also actively working on building a coordinated system of multilingual
terminology tools able to meet the many needs of the international community of
archaeologists. As part of these integration activities, the valuable work of
mapping national catalogue schemas on international standards is a critical step;
at the same time integration of terminology resources is necessary to overcome
linguistic barriers that frequently slow down the integration processes. We have
extensively described the process of CIDOC CRM encoding of the RA Schema,
released by ICCD for documenting archaeological artefacts in Italian
archaeology, in a previous work [
ICCD is the Italian Central Institute for Catalogue and Documentation, one of the seven Central Institutes of the Italian Ministry of Cultural Heritage whose main goal is to create a centralized national catalogue of Italian cultural heritage. The activity of the Institute is based on the research and development of tools, methods and standards for knowledge, protection and enhancement of the cultural and artistic heritage in Italy. It mainly provides the management of the national general catalogue of archaeological, architectural, historical, artistic and ethno-anthropological heritage, the development of cataloguing methodologies and standards, and the coordination of the technical institutions involved in the cataloguing activities on the national territory. ICCD also provides tools and best practices for implementing these standards with the clear intent of unifying and streamlining processes related with the cataloguing activities, to guarantee quality and to implement standardisation and interoperability at a national level.
To ensure that this happens e ciently, the Institute creates and releases a series of organic resources and recommendations to support the standardization process in all its aspects. These include detailed regulations describing the various tools and the way they should be used, a set of schemas and forms to collect information in a structured way according to the di erent asset types, authority les to guarantee homogeneity for the common transversal key concepts and entities, thesauri and terminological tools to provide uniform layers of information and a common language. Among the latter category, one of the most important tools released by ICCD is the RA Thesaurus, a tool providing standard names for the de nition of archaeological artefacts described using the RA schema, the ICCD standard schema used for the recording of movable objects. The RA Schema is the most used and well established standard for Italian archaeology so far. For this reason, ICCD has invested a lot of e ort in the de nition of a terminological tool able to provide standardized and unambiguous names for speci c elds of the schema. The creation of the RA thesaurus is one of the best results of this e ort. 3
In the previous work carried out together with ICCD, a detailed analysis of the
RA Schema was made to map the most signi cant model of ICCD archaeological
cataloguing system to CIDOC CRM. The RA Schema is used to record movable
objects. It is one of the most used for Italian archaeology because of the huge and
ever increasing amount of artefacts found during excavations. The RA Schema
contains a large number of descriptive information and \cross-sections" allowing
cross references with other ICCD resources. The RA Schema, together with the
RA Thesaurus, features one of the best tools of this kind in the international
panorama of cataloguing systems. The previous mapping work was carried out
on CIDOC CRM and took advantage of version 5 of the model, released in 2013.
However, in the last two years, a new version and numerous extensions of CIDOC
CRM have been released. Version 6 and the CRMarchaeo [
One of the most di cult problems to solve during the previous mapping was the representation of the \ nding" event, intended as the excavation activity during which objects are found. This event is of paramount importance in archaeology because it is fundamental to trace the object's provenance and to reconstruct its history. Following the CIDOC-CRM model, we represented the archaeological objects by using the E22 Man-Made Object class. However, to describe their relationships with the two important activities of \survey" (corresponding to the \RE" eld of the RA Schema) and \excavation" (speci ed in the \DSC" eld), CIDOC CRM core only provided a \change of ownership" relationship that hardly ts here but we decided to use it anyway. Our previous mapping appeared as shown in Figure 1.
Thanks to the release of the new extensions and a deep analysis of the
crosssection relating the RA Schema [
The RA Thesaurus di ers from the other terminological tools created by
ICCD in the very sophisticated structuring criteria it follows, made more
complicated by the large amount of information deriving from Italian archaeology
and the huge number of classi cations and nomenclatures it provides. In
particular, the thesaurus is structured according to a multilevel schema based on
concept coordination, a typical KOS activity in which concepts are combined
with each other in order to produce meaningful \sentences" that de ne complex
concepts. Generally speaking, there can be two types of concept coordination:
pre-coordination and post-coordination. The key distinction between the two
relies on when the actual coordination occurs in relation to an information retrieval
event. Pre-coordination is decided and implemented before the information
retrieval time, by a KOS maintainer or by an indexer who is using the KOS itself.
This occurs, for instance, when an indexer takes two existing concepts from a
concept scheme, such as \Coins" and \Mintage", and explicitly combines them
with a given syntax, such as \Coins-Mintage", to index a particular document.
Post-coordination, on the other hand, is performed as part of an information
retrieval task, for instance through a SPARQL query able to retrieve all documents
indexed using both \Coins" and \Mintage" concepts [
It is evident from the example above that the thesaurus itself does not o er a closed and exhaustive list of all possible terms that can be used during the compilation of the schema. Instead, it is a reference tool that, after a general term is xed, assists the user in proceeding to further speci cations by the addition of suitable subterms to gradually approximate the precise semantic meaning of the object to be described.
The exibility of this structure allows it to achieve a signi cant depth of semantics, where required, and to build speci c de nitions of several types of objects, in order to have an entry like this: Cintura per la sospensione delle armi, borchia (Belt for weapons suspension, stud) representing an exhaustive explanation of the fragmentary object itself and of the bigger object which is part of, and also as a valid entry from the terminological point of view following the formal recommendations provided by the ICCD guidelines and validation systems. 3.2
SKOS is the standard chosen by the ARIADNE project for the encoding of all terminological resources to be used in its integration plan, and for the undeniable advantages provided to integration and interoperability by its RDF-based format. As one can easily understand from what was previously stated, the \combinatorial" nature of the RA Thesaurus, and especially of the sections intended for the encoding of the OGTD eld (column four and ve), makes it very difcult to encode in a SKOS compatible format, which requires that a complete, self-consistent and self-su cient de nition exists in the thesaurus for each item or concept. The SKOS vocabulary itself does not provide any mechanism for expressing that a given concept may consist of other pre-coordinated concepts. It is, of course, possible to extend SKOS to establish a pattern for representing coordinated concepts, for instance by stating a new sub property, as in the following example: iccd:coordinationOf a rdf:Property ; rdfs:domain skos:Concept ; rdfs:range rdf:List. and then use the new property this way: iccd:coinsMintage a skos:Concept ; iccd:coordinationOf (iccd:coins iccd:mintage); skos:prefLabel \Coins-Mintage"@en.
However, patterns for pre-coordination have not yet been exploited by the SKOS community and solutions of this kind have not been explored fully enough to warrant their inclusion in the o cial SKOS vocabulary. Analyzing the RA Thesaurus, PIN and ICCU identi ed a possible solution. We tried to follow a different approach, more \pre-coordination oriented" to rearrange, where possible, the original content according to semantic criteria in order to de ne meaningful self-consistent concepts in the SKOS representation. After discussing the matter in depth, we proposed the following solutions:
1. The partitive speci cation subterms are in many cases independent terms related with the main term mostly by a part-whole relationship. Thus, it is possible to describe this relationship by using the skos:narrowerPartitive property to de ne them. This is particularly suitable if we consider that the same partitive term could occur for di erent main concepts: both a belt and a ag could have a puntale (ferrule) as partitive concept. Therefore, it is important to clearly de ne the hierarchy of these kinds of objects. Alternatively, it would be possible to combine main terms with their partitive terms in order to de ne complete and self-consistent concepts, to be then de ned as narrower terms of the main ones. In the previous example, we could de ne, for instance, a new puntale di cintura (belt ferrule) term, which would be clearly distinguished by a puntale di insegna ( ag ferrule), the two being totally di erent, although very similar, objects.
2. The morphology and functional speci cation subterms are meaningless in themselves. They become meaningful only when combined with their main term. Creating SKOS narrower terms from these elements requires, for each morphological or functional term, the creation of a subterm obtained by combination with the super concept, in order to obtain a set of semantically consistent narrow terms. There is no semantic meaning in multipla itself unless this concept is used together with cintura in order to specify, in this case, the typology of a given belt. Cintura multipla is, on the other hand, a perfectly consistent concept.
Multiple combination of partitive, morphological and functional sub concepts to create speci c entries, even if not impossible, would be very di cult to implement in SKOS due to the exponential growth of all possible combinations. At present, we propose not to extend the pre-coordination operations beyond the minimum requirements of semantic understandability and to use more than one SKOS concept to describe speci c archaeological objects if required. 4
From a technological point of view, the RA Thesaurus was created starting from 2008 on the basis of the terms extracted from the database maintained by the \Sistema Informativo Generale del Catalogo" (SIGEC). Its development also went through various phases of data cleaning and strengthening. The RA Thesaurus is currently an \open vocabulary", meaning that it is not meant to have a stable form since its content can be updated and modi ed by ICCD during further stages of research. Currently, the available version of the vocabulary is in textual format that is organized in a tabular structure, whose elds comply with the ISO standard norms for thesauri. In order to make the original textual information interoperable and ensure integration with semantic terminological tools, it was necessary to encode them in a structured and standard format.
The process we implemented for the SKOS encoding of the RA Thesaurus is a proposal for its re-engineering as a formal ontology and for making the knowledge it provides explicit in a formal sense. The whole process of encoding required a set of subsequent steps for data analysis, adjustment, conversion, publication and enrichment, in which the original textual data has been processed using both open source tools and ad hoc scripts.
The process can be subdivided into two analytic phases (see Fig. 5): 1. In the rst analytic phase we focused on encoding the key elds of the original thesaurus, such as concepts and classes. The result of the rst phase consisted in the creation of a SKOS/RDF version of the RA Thesaurus obtained through the mapping between the main concepts and the SKOS Core Vocabulary. 2. In the second phase, we focused on the integration of all morphological, functional and partitive aspects related to thesaurus' concepts. The analysis of this additional information required further investigation into how SKOS extensions could be used for the publication of thesauri in a semantic format. 4.1
The conversion of the RA Thesaurus initially required a deep data analysis to
dene a precise mapping between its main elds and the SKOS Core Vocabulary [
Figure 6 shows an example of the mapping expressed by using SKOS entities.
Each concept coming from the RA Thesaurus is represented by a blue circle.
The central circle depicts the concept of Cintura (belt) while the red circle
represents the thesaurus itself. Arrows connecting the various circles represent
the SKOS relationships existing among them. The mapping de nition on the
SKOS Core Vocabulary was followed by the use of an ad hoc script and of a
speci c tool that allowed the conversion of a huge textual le into RDF format.
At rst, the original thesaurus was manipulated and converted in order to create
a CSV le that satis ed some speci c technical requirements. The script was
developed in Perl language and was intended to select a speci c thesaurus' subset
of elds, to sort and to clean the information and to convert them into a custom
CSV le. Subsequently, the Stellar Console tool was applied to further elaborate
this le. Stellar Console is an open source command line utility application
developed in the framework of the AHRC-funded project \Semantic Technologies
Enhancing Links and Linked Data for Archaeological Resources" (STELLAR)
[
Thesaurus Publication and Enrichment Using SKOS Extensions In the second analytic phase, the publication of the thesaurus was analysed and tested on a vocabulary server. Possible solutions for mapping and integrating the elds that were not converted in the rst analytic phase were consequently studied and tested.
In order to produce the necessary results for the RA Thesaurus publication, it was important to consider two fundamental aspects. The rst was a vocabulary web server supporting international standards such as SKOS and the ISO thesaurus norms; the second was a vocabulary web application which supports multilingualism, semantic thesauri and data enrichment. All these aspects, in our opinion, are fundamental to make the RA Thesaurus even more exible for future study phases by expanding and integrating it with further multiple extensions.
We considered di erent possibilities to achieve the above-mentioned results,
by choosing TemaTres as the most pragmatic solution. TemaTres is an
opensource, web-based thesaurus management package [
The image eld of the RA Thesaurus is also a very interesting case. As
already mentioned, images increase the richness and meaningfulness of concepts,
their presence being sometimes crucial, especially in cases where proper
understanding of the archaeological objects may remain ambiguous. In a 2005 version
of the SKOS Core Guide W3C Working Draft [
Only a subset of the RA Thesaurus was taken into account to demonstrate the feasibility of these operations. The subset includes one thousand, one hundred and ninety one terms related to ten major categories (highlighted in the original source as \livello 1 categoria ") relating to: { CLOTHING AND ACCESSORIES { FURNISHING { TRANSPORTATION { CONSTRUCTION INDUSTRY { PAINTING { ARCHAEOBOTANICAL FINDINGS { ARCHAEOZOOLOGICAL FINDINGS { SCULPTURE { INSTRUMENTS - TOOLS AND OBJECTS OF USE { GENERAL TERMS
The analysis for nding the corresponding entries in the AAT thesaurus took into account the information provided by scope notes and images accompanying each concept; extensive web searches were performed to nd the most appropriate matching term between Italian and English; and terminological researches was carried out using di erent resources to identify synonyms to make the associated targetLabel as unique and as precise as possible.
The mapping work has identi ed : { 457 broadMatch associations { 104 closeMatch associations { 630 exactMatch associations
Three examples of association are provided in the following table:
At the end of the mapping work we can say that the most signi cant activity, from the scienti c-methodological point of view, has been the review of the whole process. Started as punctual control \1: 1" of correspondence between the terms of the two terminology tools (thesaurus ICCD / RA and AAT), this review has, in fact, been expanding by realizing the mapping of the terminological categories relating to individual entries with the codes referring to the facet and the hierarchy AAT. This has made possible: 1. disambiguating and correcting matches previously selected - and often lexically corrected - but decontextualized from the original domain of belonging; 2. providing the basis for a future matching job between di erent categories of multilingual thesauri.
It is worth underlining that the focus of the whole work of mapping is the concept of individual terms meant as records entered in a complete hierarchical structure of related terms and notes. Among the results which have been achieved - and which are highlighted though the mapping between classes - we can state the high level of correspondence between the ICCD/RA thesaurus entries and the AAT Thesaurus record types. Out of one thousand, one hundred and ninety one basic records one thousand, one hundred and sixty four among them are linked to \concept" and only twenty seven to \guide term". According to the AAT Thesaurus guidelines: { Concept: Refers to records in the AAT that represent concepts; records for concepts include terms, a note, and bibliography. { Guide term: Refers to records that serve as place savers to create a level in the hierarchy under which the AAT can collocate related concepts. Guide terms are not used for indexing or cataloguing. 6
The study and analysis of the RA Thesaurus allowed us to fully understand the complexity of the challenges arising from the need to de ne, by means of standard nomenclatures, objects of such various and multifaceted nature as archaeological objects are. The ICCD RA vocabulary, being the result of years of research by a team of experts in the eld of Cultural Heritage, is de nitely an irreplaceable resource that adequately meets this need. Its structure is certainly an important point of arrival on the road to standardization. From a methodological point of view, the work carried out has highlighted both conceptual and procedural challenges that arise when attempts are made to handle a complex structure in a standard tool. The results achieved so far are considered satisfactory, also in consideration of the fact that the work is at an intermediate stage and that further studies and investigations will be necessary before the conversion of the entire thesaurus can be completed. Future activities will include a clear and unambiguous de nition of complex concepts, such as those arising from the combination of multiple terms and subterms; and the de nition of precise criteria for the inclusion of images, which, as stated, is one of the distinctive features of this vocabulary. The choice of AAT as the common standard partially solves the multilingualism issues, providing labels in di erent languages for the terms already mapped. We must instead provide appropriate translations for those that have no equivalent in the thesaurus of the Getty Institute. At the end of the ARIADNE project, the RA Thesaurus will become part of the rich set of terminological tools that the project is already collecting in order to integrate them into the platform on which real interoperability will take place. The ARIADNE Portal will make this resource available and easily accessible online for external use outside of the project. The publication as Linked Open Data, also provided by the project, will guarantee its availability in other Cultural Heritage scenarios. 7
The present work has been supported by the ARIADNE project, funded by the European Commission (grant 313193) under the FP7 INFRA-2012-1.1.3 call. The authors opinion do not necessarily re ect those of the European Commission