The Sapienza Digital Library collects digital resources from the di erent University's Organizations, representing the multidisciplinary Sapienza University's community. The underlay of the metadata infrastructure was built on digital library standard metadata semantics and was used for exchanging package, between the archival systems that manages di erent services for the established digital library. The semantics adopted for the metadata infrastructure can be exploited, not only for the actual digital library services, but also for connecting the resources to the Linked Open Data Cloud through authoritative identi ers.
The paper describes a speci c aspect of the development of the Digital Library System of the Sapienza university (Sapienza Digital Library http://sdl. uniroma1.it). The approach adopted collects information, regarding the Organizations involved in the management of the digital resources' life-cycle. In order to manage the complexity of the Sapienza University's organizational framework, a work ow for building digital resources, based on the Organizational semantics, was designed at the rst stage of the project's development. The creation and the maintenance of an identi cation system, based on semantics used at national level, and mapped onto other identi cation systems, internationally used, was necessary, in order to make feasible the retrieval of relevant information in the Linked Open Data Cloud1 through an authoritative identi er. The system had been resulted essential, in the entire life-cycle of the project's development, in order to refer unambiguously to the digital resources among the project's participants. In addition it was supportive for testing and improving of the overall system's information infrastructure, for re ning the metadata structures, and for curating the data.
The semantics of the SDL metadata infrastructure were used for building selfdocumenting packages containing metadata and objects, and for exchanging packages between di erent digital repository systems. The digital repositories, 2
A. Di Iorio, M. Schaerf sharing the SDL semantics, uses the exchanging package for replicating digital resources and for distributing digital library services. 2
The Open Archival Information System (OAIS) [
The system architecture was conceived with the scope of the Long Term Digital Preservation(LTDP) of materials for the multidisciplinary community of Sapienza.
The replication of the produced OAIS IPs in di erent repositories geographically
separated, and the heterogeneity of the supporting technologies and
methodologies[
As consequence, the initial scope of building a digital library was extended, and turned toward the conception of an infrastructure for digital library, and digital preservation services.
Following this conception, the metadata infrastructure had to be agnostic about the technological platform, in order to re-use information and objects in di erent digital systems, as well as in di erent semantic contexts. Nevertheless much The digital library access through semantics 3 of the semantics, used for the values of the metadata elements, are often under the competence of the managing Organization. The semantics used if not welldocumented and structured can be an obstacle, for the automatic management of data and documents, and consequently can have have a strong impact on the long term management of the digital resources.
Under this belief, the work- ow for building digital resources was conceived for absorbing information, conveying the custody chain of the management activities performed by di erent Organizations.
In other words the overall management of a digital resource, during its creation process, is permeated by the Organization's context information, connecting the digital resource to its "real" Organizations involved in the management of its production's .
An abstract representation of the main components of the overall architecture
of the system is showed in the Figure 1. The main components can be divided in
three categories: the pre-ingestion systems preparing the digital resources, the
Digital Library Management System (DLMS)[
{ The web portal of SDL, which manages the public interface of the system. The Cataloging system and the web portal had been developed using Drupal5 that uses services managed by the DLMS. The Italian University consortium Cineca6, as technological partner of Sapienza for SDL, has developed the DLMS and the Cataloguing and the web portal systems.
Actually the repository, archiving the digital resources managed by the DLMS,
is located in Bologna (the location of the Cineca's headquarter).
The exchange of IPs between pre-ingestion systems (Massive conversion and
Cataloguing) and the DLMS, is performed between Sapienza repositories in Rome,
and the DLMS's repository located in Bologna. This preservation strategy
respects the in uencing requirements of the LTDP: the digital resources replication
in di erent repositories geographically separated, and the heterogeneity of the
supporting technologies and methodologies[
A. Di Iorio, M. Schaerf The OAIS IPs produced by the pre-ingestion systems are the exchanging packages used by the systems supporting the di erent services. Consequently the IPs produced by the pre-ingestion systems has to be self-documenting, on the base of metadata and identi cation semantics shared by the SDL systems, geographically separated.
The Sapienza University's is a complex Organization composed by 63
investigation departments, 56 libraries, 21 museums, 8 administration departments and
some research center. We have conceptually considered the Sapienza's
Organizations as Organizational units belonging to the Sapienza University.
In order to deal with the Organizational complexity of the Sapienza
University, it was deemed essential to devise a metadata infrastructure, not only based
on semantics world-wide known, but also with identi cation semantics aiming
to identify unambiguously the Sapienza's Organizational units, involved in the
work- ow production of the digital resources.
The digital library access through semantics
5
Furthermore, the long-term focus implies that the metadata infrastructure is
able to record information referring to the real evolution of the Organizational
units, that are involved in the management of the digital life-cyle of resources.
The conception of an holistic approach referring to the Organizations' custody
chain, recorded and expressed by the metadata infrastructure was based on the
two reference model cited in the section 2 [
Organizational Collection. 5
The digital resources managed by the SDL system constitute the digital
representation of the Intellectual Entities[
By the implementation point of view a DR is physically composed by the set of objects les, that together represent the OAIS Content Information, and the set of metadata represent the OAIS Preservation Description Information. 5.1
The metadata infrastructure was conceived for supporting di erent kind of DRs. The DRs can be represented in di erent formats (still and moving images, texts, 6
A. Di Iorio, M. Schaerf sounds, cartographics, etc) and can be representing di erent kind of intellectual contents (multidisciplinary knowledge). In order to manage the materials' diversity and to deliver centralized digital library services, based on the metadata, we had considered metadata standards with a su cient degree of granularity, as well as a su cient level of semantic interoperability. The analysis of the standards adopted in the digital libraries' scenario had driven to the choice of a very well known standards combination: { Metadata Objects Description Standard(MODS) which describes the intellectual contents and follows libraries semantics, derived by the MARC 21 semantics7, the pillar standard of all libraries information systems; { PREservation Metadata Implementation Strategies(PREMIS) for managing preservation metadata; { Metadata Encoding and Transmission Standard(METS)8 for wrapping together metadata belonging to the DR.
Mostly these standards, made up in combination, have covered the need of providing su cient granularity of information for the intellectual content (MODS), su cient granularity of information for the digital preservation management (PREMIS), and su cient granularity and exibility for supporting the need of managing an Organization infrastructure, using DRs variously structured (METS). Indeed, the encapsulating mechanism provided with METS has allowed not only to include other standard semantics, more relevant to speci c aims (like for example Dublin Core (DC)9 (more interoperable), or NISO Technical Metadata for Digital Still Images Standard MIX10), but also supporting the exchange of packages between the architectural components of the SDL infrastructure (see Sect.3). 5.2
The metadata infrastructure is coded in the adopted standard semantics and is organized on the DRs, that are the essential bricks, building the digital library. Both the massive conversion system, and the cataloguing system produce DRs, encoded in XML11, and conforming to the metadata standards adopted by the project (see the following Section).
The DLMS ingests DRs produced by both two pre-ingestion systems, in order
to start the management of their digital life-cycle[
The Figure 2 is a simpli ed representation of the SDL's DR. On the left is visible how the conceptual OAIS IP is generally divided into two parts: the metadata, and the content objects. On the right is represented how is physically composed 7 MARC 21 Format for Bibliographic Data, www.loc.gov/marc/bibliographic/ 8 Metadata Encoding Transmission Standard, www.loc.gov/standards/mets/ 9 The Dublin Core Metadata Initiative, dublincore.org/ 10 NISO Technical Metadata for Digital Still Images Standard, www.loc.gov/ standards/mix/ 11 Extensible Markup Language (XML), http://www.w3.org/XML/ The digital library access through semantics 7 inside of the system, as a set of di erent metadata semantics and a set of object les. Each box is labeled with the name of the related standard XML schema12 name(see Sect. 5.1). The descriptive metadata, pointed by the blue arrows, is coded into two descriptive standards. The MODS which re ects the granularity of MARC 21. The DC, which is commonly adopted in other contexts, not strictly related to the libraries world, is consequently considered more interoperable.
The inventory metadata, listing the les' names and locations, and the
structural metadata, pointed by the red arrow, are coded in the two relevant METS
sections. Both sections of metadata are connected together by METS, which is
essentially used for conveying the whole structure of the DR in the XML format.
All metadata blocks are unambiguously identi ed and referred to the
Organizational context[
The system was publicly opened the 20th of December 2013, as Beta version 1.013, and is under testing by the communities.
The DLMS is actually providing access, and discovery services to the communities and has ingested more than 11.000 DRs distributed in 22 collections, belonging to 10 di erent Organizational units. By the end of the year more than 30 new Sapienza's Library will be incrementing the number of digital resources.
The SDL DR's production starts from the main constraint of existence about the identi er of one of the Sapienza's Organizational units, which has the man12 XML schema, www.w3.org/XML/Schema 13 Sapienza Digital Library, sdl.uniroma1.it agement responsibility of the DRs. Consequently, the conditio sine qua non for the existence of every single DR must be its identi cation by an identi er based on the Sapienza's Organizational units' identi er. This identi er abstractedly de nes the concept of "Organizational collection", that gathers all DRs belonging, owned or managed by a Sapienza's Organizational unit. Consequently, all objects belonging to a DR are identi ed extending the Organizational collection identi er, which is the root of the identi cation.
The identi er is necessary for capturing information about the Organization context, which has some responsibility in the SDL DR production: scienti c or technical responsibility, objects digitization, metadata editing or management responsibility. The long term focus of the digital library requires to deal with an ever-growing amount of DRs and the re-use in the long term of a DR could result di cult or inconsistent, if it is not possible to have agents of reference about its management.
The semantics adopted for the whole process of SDL's DRs production is based on an identi cation system that, rst of all, aims to identify the Sapienza University ownership of the digital library service. In addition it identi es the Sapienza's Organizational unit, having the initial management responsibility of the resource's digital born under the Sapienza domain (selection or creation of the digital materials). The production method designed for building DRs allows to produce self-documenting IPs, where the documentation is based on the structured semantics, referring to the Organizational context. 6.1
The Organizational collection and the identi cation family The Organizational collection in the conception of the SDL is the digital embodiment of the Organization's collecting actions, that consist of the digital production, preservation and fruition. The collected digital item is represented physically by the OAIS IP which is the DR in the SDL context. The Organizational collection is the set of the whole digital production made, managed or owned by the Sapienza's Organizational unit that has the responsibility of the DRs created for the SDL. The abstract concept of the Organizational collection refers the contextual information about the Organization and set the basement of the identi cation semantics of the referred DRs.
By means of the Organizational collection identi er, we captured information
about the organizational context where the DR was born, and produced for
the ingestion in the SDL's DLMS. We have also leveraged on the identi cation
information for relating other information, about context and provenance[
This is the reason why the related Organizational collection's identi er is considered the rst mandatory information, for submitting the resources to the system.
In order to respect the LTDP requirement, allowing the DRs re-use, we have considered essential to use identi cation semantics, already used by a national identi cation system, where the main organization Sapienza and its Organizational units are hierarchically represented. The digital library access through semantics 9 Respecting the hierarchical structure of the University, the SDL identi cation system has adopted an identi ers' family derived and extended from the Italian National Bibliographic System14 where the Sapienza University is identi ed by the identi er "RMS".
This is the main identi er, which associated with descendant identi ers,
unambiguously identify the Organizational collection, and build relationships with
other entities involved in the DRs management: objects, agents, events and
rights[
The well-de ned structure of the SDL identi cation system has allowed to enrich resources and the pertaining objects with contextual information about Sapienza organization.
In addition, the registration of the Sapienza University to the international identi cation MARC organization code 15, identi ed by "itrousr", and semantically mapped to the same level of the italian "RMS" identi er, allows to set the DRs context also at international level. Indeed, the replication of such code as mandatory administrative metadata in each SDL's DR, makes possible its connection to the Linked Open Data Cloud16.
The open world "itrousr" identi er, exposed by the Library of Congress Linked Data Service(LCLOD)17 in the Cultural Heritage Organization identi cation system as authoritative identi er, makes each DR, belonging to the local Sapienza domain, worldwide reachable through the exposed identi er "http://id.loc. gov/vocabulary/organizations/itrousr", and by virtue of the mapping between the local ("RMS") and global identi er("itrousr"). 6.2
The Organization as the source of the identi cation system The SDL identi cation system is structured on four layers, extended from the main layer, represented by the "RMS" identi er of the Sapienza Digital Library, and going down to the following hierarchical layers, that are also sampled in the Fig.3: { the root identi er corresponding to the Organizational Collection (see subsect. 6.1), in the showed case, "RMSAR" identi es the Sapienza's Library of Architecture; { the digital collection identi er, corresponding to the SDL aggregation level for managing DRs, which in many cases is directly identi ed by the Organizational collection itself. In the showed case, the Library of Architecture collects the digitized books from its holdings, directly collected as DRs of the Organizational collection "RMSAR". In addition the same library collects a special collection "RMSAR SEVERATI" collecting images, donated by an Architecture's Faculty member; 14 Anagrafe Biblioteche Italiane http://anagrafe.iccu.sbn.it/opencms/opencms/ 15 MARC code list for organizations http://www.loc.gov/marc/organizations/ org-search.php 16 Linked Open Data, http://linkeddata.org 17 Library of Congress Linked Data Service, id.loc.gov
A. Di Iorio, M. Schaerf { the DR (Figure 1) identi er, in the gure the "RMSAR 00000025" is a digitized book of architecture, and "RMSAR SEVERATI 00000001" is a photograph digitized and containing Brasilian buildings relevant for the architecture interest; { the digital objects identi er, represented by the DR's identi er and the order number of the object, as example the book's page "RMSAR 00000025 0324". The replication of the higher layer's identi ers over the identi ers of the lower layers, allows to reuse the single objects in other contexts, without ambiguity about the pertaining DR of the objects, and from the root identi cation layer, back to the responsible Organization. The multiple representing format (in the example jpg and tif) are managed by the system, using the reference of the digital object's identi er.
The management of the identi ers based on semantics, derived by the Organizational collection conception, matches with two Ontologies, recommended recently by the W3C.
The Organization Ontology(Org-O), originally developed for use by data.gov.uk18, represents the formal de nition resulted from real implementations and uses. The 18 Opening up government http://data.gov.uk/ The digital library access through semantics 11 core class in the ontology is the "Organization" class which represents "a collection of people organized together into a community or other social, commercial or political structure". The main class "Organization" of the ontology Org-O, semantically speaking, matches to the Sapienza University. While the Org-O subclass "OrganizationalUnit", matches with the Sapienza's Organizational units. The matching conceptualization between the "OrganizationalUnits" class of the Org-O and Sapienza's Organizational units associated to the SDL's Organizational Collections, and unambigously identi ed, will drive the reasoning systems to retrieve information about DRs belonging to the pertaining "Organization" or "OrganizationalUnit".
The identi cation system based on semantics locally de ned, but world wide processable by means of dereferenceable URI like the LCLOC identi er (see Sect.6.1), allows to make all belonging DRs reachable by URI through the Organization ontology support.
Coherent to this scenario is the ontology aimed to model the information about "entities, activities, and people involved in producing a piece of data or thing, which can be used to form assessments about its quality, reliability or trustworthiness"19, known as Provenance Ontology.
The Prov-O Ontology(Prov-O) [
The recommendation by W3C of this two ontologies demonstrates the global interest, around the traceability of digital assets back to the Agents responsible for their management, harmonically with the SDL's Organizational collection conception, where the agents belong to the context information referred to the Organization.
A. Di Iorio, M. Schaerf