In this paper we describe the pilot project dati.culturaitalia.it, which started in 2012 to build up a Linked Open Data (LOD) Service that will progressively make available open datasets from the web-portal CulturaItalia 1, the Italian national aggregator for Europeana 2. CIDOCCRM Ontology was used for transformation and representation of data widely pertaining to the cultural domain. RDF triples mapped into Erlangen CRM were then enriched with links to URIs identifying instances of internationally established RDF resources for geographic names, and instances of authority les for personal and corporate names, such as GeoNames and Virtual International Authority File (VIAF). CulturaItalia is the Portal of Italian Culture, promoted by the Italian Ministry of cultural heritage, activities and tourism (MiBACT), in which cultural institutions from all sectors and levels (national, regional and local) are involved. CulturaItalia also plays an important role for the development of Europeana, making available cooperative networks and agreements and coordinating technical activities leading to the establishment of Europeana environment.
CulturaItalia [
gathers the metadata to Europeana, the public digital library promoted by the European Community. CulturaItalia is targeted to general users, by o ering them a service for retrieving information on Italian culture from one accesspoint, and to more expert users, such as the operators in the cultural eld, who can take advantage of a high-quality showcase to promote their own digital resources. CulturaItalia makes the digital resources interoperability possible, through a cross-domain Application Pro le (PICO AP: PICO is the acronym for \Portale della Cultura Italiana On-line"), based on the Dublin Core Metadata Initiative technical guidelines. The Portal gives access to a rich \metadata" collection, which gathers and organizes information arriving from the various providers participating in the project. Users can discover di erent kinds of digital resources, describing the country's extensive cultural heritage (museums, photographs, libraries, archives, galleries, exhibitions, monuments, audio-visual works, etc.). The pilot project dati.culturaitalia.it started in 2012 with the aim to build up a Linked Open Data (LOD) Service that will progressively make available open datasets from the Web-portal. The application was designed by the CulturaItalia team with the technical and scienti c support of Scuola Normale Superiore, and was developed by Meta s.r.l., to allow the resources aggregated by CulturaItalia to be involved into large semantic networks after exposing, sharing and connecting data according to LOD principles. A rst release of this service is available on-line since 2013 3 as a section, or sub-portal, of CulturaItalia dedicated to LOD. It presently makes available as LOD the Thesaurus PICO, adopted by the portal for facilitating the browsing of a variety of resources in its domain, and a selection of metadata sets from the Portal. The CulturaItalia team has chosen CIDOC-CRM, in the implementation of Erlangen CRM/OWL, to foster the interoperability in the cultural heritage sector. In the perspective of a future integration with the bibliographic heritage of the Open Catalogue of the National Librarian System (OPAC SBN), managed by ICCU, the Institute implemented, in 2014, a mapping activity, with the support of a team from VAST-LAB (PIN), to convert resources from OPAC SBN, encoded in UNIMARC format, in FRBRoo, adopting the CIDOC-CRM model. 2
CulturaItalia manages a catalog - called Index - which gathers and indexes metadata provided by the partners. The original data remain on the Web-site of the provider, to which the nal user is redirected by CulturaItalia, through links, thus allowing to retrieve the original and complete information. For example, in the case of a photograph, in the CulturaItalia Index the preview image (thumbnail) is visible, together with some identifying data, and a link to the provider's website allows the user to visualize the photograph in its original format, accompanied by the complete information and services, in order to get the full bene t of the item. The resources in the Index are classi ed on the basis of the PICO Thesaurus, designed to manage and organize heterogeneous information, from
di erent cataloging systems. Browsing the Index, the user consults the metadata through a hierarchical classi cation of terms (facets). CulturaItalia is an \open" system: it grows up and develops together with the continuous enrichment of its metadata Index, through the metadata harvesting according to OAI-PMH, a protocol which allows the harvesting of metadata from content providers to one or more harvesters, adding services as indexing system or automatic classi cation. The Portal harvests metadata from di erent repositories and exports metadata to other national and international portals and repositories. At present CulturaItalia aggregates over 3 million metadata from 32 public and private partners including thematic aggregators, such as Internet Culturale, the portal of Italian Libraries, also created and managed by ICCU. Internet Culturale plays a keyrole in guiding the libraries in the production of standardized digital cultural resources and metadata, according to the Italian standards. Metadata published in Internet Culturale are automatically transferred to CulturaItalia, and then, if the providing libraries agree, to Europeana.
All content from external data sources are integrated in CulturaItalia in the form of metadata, thanks to the OAI-PMH harvesting protocol and published in the portal using a speci c Application Pro le (PICO AP), based on the international standard language, Dublin Core, that can describe, in a single scheme, every type of cultural resource, both physical and digital. Following DCMI recommendations, Scuola Normale Superiore di Pisa (SNS), which supports the ICCU working group engaged in the development of CulturaItalia, de ned an application pro le which joins DC Element set, Quali ed DC terms and some further re nements and encoding schemes conceived for the application of CulturaItalia. The PICO AP combines in one metadata schema all DC Elements, all DC Element Re nements and Encoding Schemes from the Qualied DC, and other re nements and encoding schemes speci cally conceived to retrieve information pertaining to Italian culture. This Application Pro le could be further expanded for harvesting possible unexpected contents in the future, by adding Re nements and Encoding Schemes that could be suitable for data retrieval. The PICO AP can be consulted at: http://purl.org/pico/picoap1.0.xml. Schemas used for the PICO AP are published on a PURL, under the domain PICO: http://purl.org/pico/1.1/pico.xsd and http://purl.org/pico/1. 1/picotype.xsd. One of the most relevant encoding schemes introduced in the PICO AP is a Thesaurus specially conceived for the project itself, which comprehends hierarchically structured keywords indicating the topic of all the resources included into CulturaItalia (PICO Thesaurus 4.3). This ontology is also used to support the browsing into the Index of resources of Culturaitalia, therefore the assignment of a value taken from the PICO Thesaurus is mandatory for each metadata record. During the metadata generation, this assignment can be created for a whole repository or for a whole set, while in some other cases it was necessary to interpret a given value of the original database in order to create a mapping into the Thesaurus. The PICO Thesaurus is organized in four main categories: \Who" includes both people and corporate bodies; \What" comprehends tangible and intangible heritage, and all digital objects; \Where" covers Italian places (from regions to towns and villages) and \When" includes a list of chronological keywords associated to a sharp range of years. In order to be more compliant with international best practices, it seemed useful to adopt a SKOS format for the PICO Thesaurus. The SKOS format for Thesaurus PICO has also been designed to be extended and/or integrated with di erent thesauri pertaining to speci c domains, managed by institutions that have a role in standardization, such as ICCD and ICCU, or to support multilingualism through the mapping between di erent national KOS. 3
PICO AP is a Dublin Core Application Pro le. As already pointed out in the
literature related to mapping between Dublin Core and CIDOC-CRM, for every
value of the DC element \Type", specifying a type of a described resource, it
must be speci ed a di erent mapping to a main entity of CIDOC CRM [
E.g.: IF DCMITipe = Image, THEN the described resource must be mapped
as CIDOC-CRM entity = E38 Image. Consequently, each record encoded
according to PICO AP will produce one main CIDOC-CRM corresponding entity,
and the mapping of all the other PICO AP elements describing the resource
will depend on the high-level mapping between the type of the resource and the
corresponding CIDOC-CRM entity. Within PICO AP, dc:type element is
mandatory and repeatable (occurrence: min 1, max unbounded) [
From this high-level mapping, based on the type of the described resource, derive di erent mappings between the various types of PICO AP resources and a corresponding CIDOC-CRM entity.
For a PICO resource with DCMIType= \PhysicalObject" (= crm:E22 Man Made Object), the PICO AP element < pico : author > must be mapped as shown in gure 2.
For a PICO resource with DCMIType= \StillImage" (= crm:E38 Image), the same PICO AP element < pico : author > will be mapped as shown in gure 3.
As CRM entities are ordered in a poly-hierarchy, and as the properties associated to each class are inherited by the subclasses, it is possible to group the 12 identi ed entities (and, consequently, the mapping to be implemented) into 4 main groups: 1. E18 Physical Thing contains the mappings for: E22 Man Made Object and
E78 Collection 2. E28 Conceptual Object contains the mappings for: E73 Information Object,
E29 Design or Procedure, E33 Linguistic Object, E36 Visual Item, E38 Image 3. E39 Actor contains the mappings for: E40 Legal Body, E21 Person 4. E5 Event
The detailed mapping containing four mapping tables (one for each of the
above-listed CRM main entities) is available on-line within the document:
\Mapping between PICO Application pro le and CIDOC Conceptual Reference Model"
[
The SPARQL endpoint provides access to RDF metadata structured according to the CIDOC - Conceptual Reference Model in the implementation of Erlangen CRM/OWL. Data can be searched over three querying interfaces, corresponding to three sections of dati.culturaitalia.it: { Text search: here it is possible to perform free text searches over all triples contained in dati.culturaitalia.it. { SPARQL query: here you can try your hand at a SPARQL query. There are also some examples of queries. { iSPARQL query: here there is an even more complex querying interface for advanced users.
Dati.culturaItalia.it exposes an OAI Provider that makes available XML or
RDF metadata structured according to di erent schemas:
{ oai-dc (xml): OAI-PMH schema adopted by Open Archives Initiative
Protocol for Metadata Harvesting
{ pico (xml): PICO Application Pro le, the CulturaItalia Application Pro le
{ edm (rdf): Europeana Data Model, adopted by the portal Europeana EDM
[
Erlangen CRM / OWL 5
Mapping between UNIMARC Bibliographic Format / SBN MARC and FRBRoo and next steps ICCU is moving another step towards the Italian Linked Cultural Data Cloud by starting the mapping study of data from the OPAC SBN (On line Public Access Catalog of National Library Service) in UNIMARC format to the class and the properties of FRBRoo, on the base of the model CIDOC CRM. The collective catalogue of National Library Service provides access to 13.759.767 bibliographic records that contains: { descriptions of documents acquired from SBN libraries starting from the '90s or since single libraries entered the SBN { descriptions \book in hand" of documents of XVI - XX centuries { descriptions obtained from catalogues on paper previous to 1990
In 2014 a working group formed by experts from ICCU and VAST-LAB (PIN) was established with the objective to analyze and test the publication of a subset of signi cant data in UNIMARC format as LOD according to the document FRBR object-oriented de nition and mapping to FRBR-ER (version 0.9). In particular, this activity focused on: { analyzing and de ning a basic methodology for creating Linked Open Data from bibliographic archives according to the international standards and cataloguing rules adopted by SBN; { designing a schema with the conceptual description of how to relate SBN bibliographic information in a semantic way. FRBRoo, an harmonization between FRBR original conceptual model and CIDOC CRM, has been chosen as the reference intellectual guide for this activity { selecting a rst set of bibliographic records to be exported from OPAC SBN in UNIMARC format { de ning all the required namespaces and URI mechanisms to create meaningful identi ers for the converted UNIMARC entities
Activities performed by the working team lead to the de nition of a mapping document describing the conceptual mapping between UNIMARC elds and FRBRoo entities and properties, with speci c de nition of mapping paths for every possible combination or special use cases of UNIMARC encoded information available in the SBN archive. The selected records subset was used for testing the conceptual coherence of the model in order to identify possible con icts and to x co-reference and cross-reference issues that might have arose.
Speci c exporting scripts have been developed to encode the UNIMARC bibliographic information in a standard RDF format, to transform it in a machinereadable version using a formal language. Bibliographic information created in this way was afterward enriched with entities coming from VIAF, GeoNames, Linked Heritage, DBPedia, and other available online Linked Open Data resources.
A web tool has also been created to store semantic records and to query and retrieve relevant bibliographic data according with given semantic criteria. The tool is composed of various modules e ciently interacting with each other and based on open source technology. The modules include: { an online triple store based on Sesame to accommodate the RDF triples created by the exporting framework and to manage the complex network of relationships de ned by means of it; { a set of responsive web interfaces based on Ajax/JQuery technologies and implementing the various features of semantic query and presentation of the relevant results. A basic faceted system for a more e cient browsing of the results was also implemented within the same interfaces.
The web tool also o ers the possibility to download the full Linked Open Data network of bibliographic information in an RDF compatible format for local use. Further work on this topic will necessarily require a data cleaning phase for consolidating the legacy database in order to create a better representation of its content during the mapping and conversion process. Additional activities will concern the creation of a SPARQL end point for advanced semantic queries, the improvement of the web interface to allow connection of various libraries to the SBN index, to facilities retrieving and FRBRoo encoded triples representing entities of interest (work, expression, etc.) and the export of the same information in a standard Linked Open Data format for them to be used by other bibliographic tools and in other similar contexts. Data validation cycles to ensure the full compatibility of the formats with the fundamental principles of Linked Open Data and Semantic Web philosophy will be also performed, as well as multiple tests on the internal coherence of the newly created dataset.