In this paper, we present the ArchaeoTerm Project, along with one of its first efforts in enhancing multilingual access to Archaeological data, making available a resource of Archaeological terms within the framework of YourTerm CULT project. In order to enhance and promote the use of a terminological common ground across different languages the Archaeo-Term multilingual Glossary is intended both for scholars, experts in the field, translators and the general public. Its first release contains terms in Italian, English, German, Spanish and Dutch together with PoS, definitions and other linguistic information. This paper presents the data and the methodology adopted to create the glossary as well as the evaluation of the first results.
Languages for Special Purposes (LSP) have their roots in the need of communicating specialised and technical knowledge within a restricted group of domain experts.
From a linguistic perspective, LSP are mainly
characterised by the use of specialised
terminology, which is usually monosemous for the
principle of clearly defining concepts and avoiding
miscommunication and can often result opaque and
unintelligible to laypeople
Copyright c 2020 for this paper by its authors. Use
permitted under Creative Commons License Attribution 4.0
International (CC BY 4.0).
dactic and informative functions of
communication
The language used in the domain of Cultural Heritage (CH), and its sub-domains, such as Archaeology, shares many points with other LSPs, such as the presence of technical terminology, terms of Greek and Latin origins, re-semantisation of common words into specialised domains of knowledge, complex multiword expressions, to mention a few. Nonetheless, it has been traditionally less investigated if compared to, for example, the language of medicine or law, which are considered soft disciplines too. As a consequence, except for a few felicitous examples (see Section 2), language resources and especially terminological resources, in this domain, are still needed. Language resources such as glossaries, thesauri, dictionaries and term-banks are invaluable sources for language experts, translators, learners, among others. Their development can often be demanding and time-consuming, especially when carried out manually.
Specialised domain resources are even more challenging because their creation also needs the validation of experts in the domain of knowledge. In this paper we present our work aimed at the creation of a multilingual glossary of archaeological terms, which is useful in many application scenarios from Machine Translation (MT) to Natural Language Processing (NLP).
The remainder of the paper is organized as follows: Section 2 describes related work and, following this, Section 3 presents the Archaeo-Term Project’s aims and the creation of the multilingual glossary of archaeological terms, along with the description of the starting data used so far, namely the ICCD Thesaurus, and the methodology applied to extract multilingual data from the Getty AAT. To complete this section, we illustrate the first results together with their evaluation. Finally, the paper ends with the conclusions and the future 2
Terminology, as several scholars pointed out
More recently, with the spreading of the
Semantic Web Technologies, many language resources
are being released in compliance with the Linked
Open Data (LOD) principles, using formalisms
such as SKOS and Ontolex-Lemon, which are
based on the Resource Description Framework
(RDF), for representing glossaries, vocabularies
and taxonomies
In the field of CH some language resources have been released during the years, both monolingual and multilingual. Among the multilingual resources, the most referred one in this domain is the Art & Architecture Thesaurus (AAT)2, developed and maintained by The Getty Research Institute. It is a multilingual thesaurus used to describe art, architecture, decorative arts, material culture, and archival materials, which can be accessed through a web interface or via its LOD version (JSON, RDF, N3/Turtle, N-Triples), as well as XML and relational tables.
Another multilingual terminological project on CH is the iDAI.vocab3, a controlled vocabulary specifically designed for archaeological terms available in several languages, developed by the German Archaeological Institute (DAI). Many other glossaries and thesauri have been created as monolingual resources for cataloguing purposes. Such as the vocabularies developed by the FISH (Forum on Information Standards in Heritage)4 and maintained as LOD resources by the Heritage Data5 for English, or the thesauri and controlled vocabularies developed by the Italian Institute for Cataloguing (Istituto Centrale per Il 2https://www.getty.edu/research/tools/ vocabularies/aat/about.html
3https://archwort.dainst.org/it/vocab/ index.php
4http://www.heritage-standards.org.uk/ terminology/
5https://www.heritagedata.org/blog/
vocabularies-provided/
Catalogo e La Documentazione - ICCD) 6.
The ICCD has also started, in 2017, the ArCo
project7 together with l’Istituto di Scienze e
Tecnologie della Cognizione (ISTC) del CNR, in
order to make available data from the General
Catalogue of Cultural Heritage according to the LOD
principles
Some glossaries are also released by the museums or cultural institutions such as the British Museum’s Object Names Thesaurus8.
In the field of Cultural Heritage in general, and
particularly, in archaeology, it is worth
mentioning the ARIADNE Project
The Archaeo-Term project of the UNIOR NLP Research Group9 of the University of Naples “L’Orientale” is part of the YourTerm CULT initiative10 in partnership with the Terminology Without Borders program fostered by the Terminology Coordination Unit (TermCoord)11 of the European Parliament’s Directorate-General for Translation (DG TRAD). Among the different projects, YourTerm CULT is specifically designed to operate in all aspects of culture.
The Archaeo-Term project has been launched to fill the gap in an important field which takes us back to the roots of European culture and history, namely Archaeology.
The project aims at improving the accessibility of the archaeological information available in various sources (scientific papers, texts addressed to general audiences, web sites, structured databases, etc.) by creating language resources useful to NLP and MT tasks across languages. This will ease the availability of the information that can be used to structure and connect different types of knowledge bases together, both structured databases and un6http://www.iccd.beniculturali.it/it/ strumenti-terminologici
7http://stlab.istc.cnr.it/stlab/ project/arco/
8http://terminology.collectionstrust. org.uk/British-Museum-objects/
9https://sites.google.com/view/ unior-nlp-research-group 10https://yourterm.org/yourterm-cult/ 11https://termcoord.eu/ structured text collections.
Indeed, although some scientific communities felt the need to structure their knowledge by means of thesauri or ontologies, the scenario is still very fragmented as posed by Felicetti et al. (2018). Nowadays, European archaeological documentation consists of a multifaceted series of information, produced in different and independent ways by each of the various national and international institutions active in this discipline, by means of tools and methods that are often very different from each other. Thus, there is still the need to establish a terminological common core shared across languages.
In this scenario, the Archaeo-Term project tries to contribute to the improvement of scientific cooperation and advancements by attracting both academia and museums from different countries in the creation of a wide multilingual terminological resource in Archaeology. With this aim in mind, one of the first results of this project is a multilingual Glossary of archaeological terms which is mainly useful for the multilingual digitalisation efforts of the museums, but also to scholars, translators and the general public. 3.1
For the creation of the Archaeo-Term multilingual glossary, we start from the RDF/SKOS version of the Italian ICCD Thesaurus12, one of the best practices adopted by the Italian Ministry of Cultural Heritage (MiBAC) to publish institutional information as LOD, in order to be easily findable, reused and freely shared. It contains 1,059 Italian terms which are linked to the LOD version of the Getty AAT13, by means of the skos:closeMatch property pointing to the Getty URIs (Figure 1). This property is used to link two similar concepts that can be used interchangeably in some information retrieval applications (Cfr. SKOS Recommendation 18 August 2009). We choose to extract the information stored into the Getty AAT because it is a valuable and trustworthy resource, created by experts in the field.
12https://github.com/ICCD-MiBACT/ Standard\-catalografici/blob/ master/strumenti-terminologici/beni\ %20archeologici/ICCD\_Thesaurus\ _definizione\%20del\%20bene\_reperti\ %20archeologici.rdf
13For the mapping process see the ARIADNE project described in Felicetti et al. (2015) The exploitation of the ICCD resource to read URIs pointing to Getty AAT contributes to build our multilingual glossary of archaeological terms along with the corresponding definitions and sources in other languages, namely English, Spanish, German and Dutch. Among the many languages available in the Getty AAT, we decide to use for our glossary those mentioned above since they show the best coverage in terms of linguistic equivalence (translations) starting from the Italian terms in the ICCD thesaurus.
In order to perform this, we use the Getty AAT SPARQL Endpoint14 to access term related information by means of setting queries. In detail, the querying process consists of a matching operation between the results of integrated queries in the AAT SPARQL Endpoint.
We first use a query capable of parsing the ICCD resource and reading each URI which refers to the corresponding English archaeological term. In fact, in the Getty AAT, English terms and other available corresponding terms in different languages are represented as equivalent terms by means of the skos:prefLabel property15 and as alternative terms in skos:altLabel property16. Both properties carry one lexical value and one language tag, associated with the lexical value, for each URI.
Since we try to extract corresponding terms in different languages, we then perform a further query able to extract archaeological equivalent terms along with their language tags and alternative terms along with language tags for each available language per URIs.
In addition to this, we set another query able to read URIs and collect corresponding definitions and sources along with their language tags, (both contained in the skos:scopeNote property)17. As a first result of such a query looping over ICCD URIs, we collect archaeological terms, definitions and sources. These queries guarantee the exploitation of the Getty AAT resource but, regardless of the language tags, also a combination of each term value associated with each definition and source value (present in the skos:scopeNote). 14http://vocab.getty.edu/sparql 15https://www.w3.org/2012/09/odrl/ semantic/draft/doco/skos_prefLabel.html 16https://www.w3.org/2012/09/odrl/ semantic/draft/doco/skos_altLabel.html 17https://www.w3.org/2012/09/odrl/ semantic/draft/doco/skos\_scopeNote.html To the best of our knowledge, in the AAT we did not find a direct link between the different language terms values (stored in skos:prefLabel and skos:altLabel and the different language literal values (definitions and sources in skos:scopeNote) represented for the same URI. Therefore, to build our multilingual glossary we rely on a matching operation between URIs and language tags related to term values (represented in skos:prefLabel and skos:altLabel), definitions and sources (both represented in skos:scopeNote). In particular, starting from a combination of all term values and literal values (definitions and sources) per language present for an URI, we apply a matching operation able to select only the terms, definitions and sources concerning the same language based on the reference URI. This matching operation allows us to recognise and organise archaeological terms and their literal values, that is definitions and sources, pertaining to the same language for each archaeological term identified by URI. 3.2
Once the queries steps are performed, we first replace retrieved URIs with numeric IDs in order to provide an identification code for each entry of our glossary; then we build monolingual tables for each language mentioned above and a multilingual synoptic table.
For monolingual tables, we automatically classify in separated tables all retrieved data based on the language tag for each term entry. On the other hand, we align the terms in the different languages based on the shared ID to build the multilingual synoptic table.
In detail, the Glossary first release18 is organised as follows:
For each language forseen in the glossary (Italian, English, Spanish, German and Dutch) there is a dedicated monolingual table, named after the corresponding language locale (e.g., IT for Italian, EN for English) which contains 8 fields (ID, Singular Term, Plural Term, Qualifier19, PoS, Alternative 18https://drive.google.com/file/d/ 1cKvZPd6bdh7lrZ6plj1gGKatWvopqFo4/view The Glossary is released under Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) 19The ‘Qualifier’ field, enclosed between brackets, indiTerms, Alternative Terms Qualifier, Definition and Source) as shown in figure 2. a multilingual synoptic table contains all the languages singular terms, which are linked to one another by means of the IDs. This multilingual table aims at providing a comprehensive overview on the equivalent terms across the languages.
During the evaluation phase, we noticed that 9 Italian terms had two equivalent English terms in the Getty AAT, marked by two closeMatch URIs to the AAT instead of just one.
A manual evaluation revealed that one URI leads to a more generic term and the other one to a more specific term. For example the Italian term letto is linked both to the Getty AAT ‘Bed’ (generic) and to ‘Canopy Bed’ (specific). In these cases, instead of following the URI pointing to the specific reference, we choose to follow the most generic one, in accordance with the Italian term meaning. We opt for a manual evaluation due to the low presence of this phenomenon, but, alternatively, it could have been performed automatically making use of an external resource such as a dedicated dictionary. Furthermore, the evaluation phase revealed a difference in the granularity of terms between the Italian ICCD Thesaurus and the other languages coming from the Getty AAT. Indeed, while the Italian terms result to be highly specific and fine-grained, many equivalents in the other languages are more in a relation of hyperonymity/hyponymity. For example, in the Italian Thesaurus there are several semantically and linguistically different types of relieves: their meanings change according to the following adjectives (e.g., Rilievo + storico, funerario, votivo, could be in English historical, funerary, votive + Relief). Nonetheless, the retrieved equivalent in English extracted from the Getty AAT is always ‘Relief’, as well as in Spanish is always ‘Relieve’ and in Dutch is ‘Relie¨f’.
Finally, some terms in the different languages, as well as some definitions, are missing and we plan to implement the missing fields in the future. Table 1 shows the total number of terms for each language in the terminological database. Missing fields are due to data sparsity, since for each Italian term there are not always equivalent terms in cates the subfield the term belongs to, thus allowing the disambiguation in case of homographs (e.g. Ax (weapon) vs. Ax (tool))
In this paper we present our Archaeo-Term Project aimed at the creation of a multilingual glossary on archaeology. The Glossary is the result of an extraction and merging process from two already available resources released according to the RDF Data Model, namely the RDF/SKOS version of the Italian ICCD Thesaurus and the LOD version of the multilingual Getty AAT.
The Archaeo-Term glossary is an ongoing project which will address, as future steps, the completion of missing data (terms, definitions, correspondences, examples, etc.) for English, Dutch, Spanish and German, as well as the enlargement of the glossary on the basis of the semi-automatic extraction of terminology from specialised corpora and other existing glossaries for the languages currently foreseen.
Furthermore, we also plan to implement the glossary with other languages such as French, Swedish, Chinese and Russian.
As future work we also plan to convert the result
of Archaeo-Term project into more formalised
formats, i.e., both TBX format (TermBase eXchange)
to be used in connection with CAT-Tools and
Ontolex-Lemon Model
Finally, when we achieve a more complete version of the glossary we plan to publish it also on a Research Infrastructure Repository such as CLARIN.
This work has been partially supported by Programma Operativo Nazionale Ricerca e Innovazione 2014-2020 - Fondo Sociale Europeo, Azione I.2 “Attrazione e Mobilita` Internazionale dei Ricercatori” Avviso D.D. n 407 del 27/02/2018 and by POR Campania FSE 2014-2020 “Dottorati di Ricerca a Caratterizzazione Industriale”. We would like to thank Michele Stefanile for his support as expert in the domain of Archaeology. Authorship Attribution is as follows: Giulia Speranza is author of Section 2 and 3.2, Raffaele Manna is author of Section 3.1, Maria Pia di Buono is author of Section 1 and Johanna Monti is author of Section 3 and 4.