Bibliographic Data Towards the Semantic Web: A Review of Key Issues and Recent Experiences

This article intends to review the underlying concepts and technologies of the Semantic Web and the potential they provide for metadata management covering bibliographic resources. To get closer to a semantic web data space, different libraries are adhering to the initiatives making their traditional Knowledge Organization Systems (KOS) operational on the web through SKOS techniques, as well as releasing bibliographic data under open licenses (open bibliographic data) and publishing it with Linked Data (LD) mechanisms. LD meaningful semantic connections create the Web of Data, a global database representing the first practical step to the Semantic Web. Here interoperable data can be processed independently of application, platform or domain, providing rich retrieval results produced by powerful query languages. From a library perspective, a problem statement is a global promotion within the Library community of understanding and of adoption of Linked Open Data (LOD), of LODe-BD recommendations, as well as releasing bibliographic data as Linked Library Data (LLD). In this way, different bibliographic datasets could become full members of the Semantic Web making interoperable different knowledge datasets of heterogeneous web communities.

Anahtar sözcükler: Semantik web, Metadata, Bibliyografik veriler, Bilgi sistemleri, Bağlantılı açık veriler BİLGİ DÜNYASI, 2012, 13 (1) 17- 56  Iryna SOLODOVNIK The Semantic Web technologies, Linked Data and SKOS The Semantic Web* (or the intelligent Web, in reference to Web 3.0) is not a separate web, but an extension of the current web which contains the virtual realm of virtually boundless information in the form of web documents.Semantic web technologies are potential enough to make logical connections of and decisions (through inference rules) on different pieces of web data, fusing (Bleiholder et al., 2008) their meaning (semantics, ontology, "shared conceptualization" 1 ), and enabling computers and humans to work better in cooperation (Berners-Lee, 1998).Indeed, the Semantic Web amplifies the conceptualization of (meta)data 2 allowing it to become semantic entities responsible to organize, access, retrieve and preserve digital information resources.
While "conceptually old for library and information professionals, metadata 3 has taken a more significant and paramount role than ever before and is considered as the golden key for the next evolution of the web in the form of Semantic Web" (Safari, 2004, p.1;  Karen, 2010).Semantic (meta)data contribute to semantic interoperability (Tolk et al.,  2007) and cross-searching of web contents.
The Semantic Web itself requires adding semantic metadata on the top of (meta) data describing web resources.This approach is aiming at processing effectively the data based on the semantic information associated with data.In this way, computers can make inferences about the data, 'understanding' what data resources are and how they relate to other data."The Semantic Web provides a framework for making data more accessible and easier to harmonize.It has the potential to unlock information that would be difficult to uncover using traditional data technologies" 4 .It is only a matter of fact to get exponentially available data (already included in Web sites, other databases, XML documents, and other systems) into a uniform format such as RDF (Recourse Description Framework).Another step is to classify and to connect data considering its properties and its relationships with other data.This is also where Semantic Web technologies such as, RDFS (Recourse Description Framework Schema) and OWL (Ontology Web Language) come in.
Summarily, the Semantic Web, whose main purpose is to create a collaborative data infrastructure 5 where to generate and exchange new knowledge, is aiming at: a. enriching information resources available in various forms on the Web through semantic annotations 6 (Rusu et al., 2011), RDF crosswalks 7 and formal descriptions/ ontologies (or web vocabularies, taxonomies 8 capturing semantics of metadata within schema structures) 9 (Valkeapää et.al., 2007); b. providing meaningful semantic connections by mechanisms of Linked Data (Heath et al., 2008; Baker, 2010), which is "the first practical expression of the Semantic Web") 10 .Linked Data infer new levels of knowledge within a global space of information resources, and also reveal new information behaviors and needs of different communities of users across the web.Links among different resources in the Linked Data Web will enable discovery of semantically related resources; c. creating a common agreed framework (through common ontology) of Cloud platforms11 .These platforms enable the exchange of information in an unified manner, and enhance the interaction of services and tasks performed by computers within a distributed network community (Kim et al., 2010; Mitchel,  2010).Through Cloud platforms, the consumer and producer agents can reach a shared understanding by exchanging ontologies, which provide the vocabulary needed for "discussion".
The architecture of the Semantic Web provides the following basic technological components: 1. URI (Uniform Resource Identifier, "Globally Unique Identifier"12 ).The URI is a formalism used to identify uniquely an information resource on the web.Each resource (single 5 To get a quick idea about a collaborative data infrastructure (built on research data), see the report "A Surfboard for Riding the Wave Towards a four country action programme on research data", <http://www.knowledge-exchange. info/Default.aspx?ID=469> 6 See TEXTUS, that has an extensible model for semantic annotations, <http://textusproject.org/>7 Semantic Interoperability of Metadata and Information in unLike Environments (SIMILE) Project.MIT, 2008."RDFizers-SIMILE" Cambridge Mass.: MIT, at <http://simile.mit.edu/wiki/RDFizers> 8 Taxonomy and a set of inference rules represent the most typical kind of ontology for the Web."The taxonomy defines classes of objects and relations among them.For example, an address may be defined as a type of location, and city codes may be defined to apply only to locations, and so on.Classes, subclasses and relations among entities are a very powerful tool for Web use" (Berners-Lee et al., 2001).9 Ontologies can improve the accuracy of Web searches: the search program can look for only those pages that refer to a precise concept instead of all the ones using ambiguous keywords.More sophisticated applications will use ontologies to relate the information on a page to the associated knowledge structures and inference rules.Here is an example of application for e-learning based on ontological structures: <http://www.merlot.org>(if different materials are organized into units -learning objects -each unit can be connected to others and reassembled in a new course).For an overview of possible applications based on ontologies: <http://www.netcrucible.com/semantic.html>;Maedche (2002).10 < http://linkeddata.org/>document, its parts and its metadata13 , objects and entities mentioned in them, image, service, e-mail) must have its URI, which can be a Web address (URL, Uniform Resource Locators, the most common type of URI) or a namespace (URN) (Berners-Lee et al., 2005).The Semantic Web, in naming every concept simply by a URI, allows anyone express new invented concepts with minimal effort.
2. RDF (Recourse Description Framework) 14 .The RDF is a W3C standard representing a declarative meta-language based on a XML-based model for encoding, exchange and reuse of metadata and their patters on the Semantic Web.The RDF provides itself the RDF data model based on the 3-part statement (triple): Resource (Subject), Property/relation (Predicate), Value (Object) (Figure 1).The automatic data triplification, according to this exposure, is closely related to human way of thinking and building concepts.

Figure 1. RDF graph of triples 15
In Figure 1, the sets of linked triples are shown as a graph of nodes and connectors identified by URIs.It is possible to imagine RDF triples connecting different data like hypertext links connecting a set of document on the web.RDF triples, which can be written with XML tags, specify relationships between "Subjects" and "Objects", in order to navigate between them.This approach provides the integration of information from multiple resources, as well as allows the fluent automatic access to different related data despite their diversified origin.
The nodes "Object", represented by rectangular, contain data that can be both literal and URI.These nodes form terminators of linked data chains because they cannot be matched to other nodes without ambiguity" (Dunsire, 2012).Nodes of "Subjects" and "Predicates" are identified by URIs, and can be processed only by machines.Summarily, the unifying logical syntax of RDF triples enables different concepts defined by URIs to be progressively linked into a universal Web, and to make logical assertions based on the associations between "Subjects" and "Objects", thus generating automatically web statements about resources.The inference among RDF predicates is possible through 3. Inference engines (web agents) or computer programs capable of interpreting RDF and OWL semantic information.They are essential component in the generation of new knowledge on the web.Indeed, the potential of the Semantic Web would be useless if there were no such inference engines gathering information from diverse sources, as well as processing and exchange it with other programs, inferring new data.
While RDF model provides a good syntax for web resources, it does not specify their semantics.For this reason, the Semantic Web offers the already cited technologies such as RDFS and OWL.
4. The RDFS 16 represents a vocabulary describing groups of related RDF resources together with their relationships.Particularly, an RDFS vocabulary expresses the acceptable properties and their values that may be assigned to RDF resources within a given domain.Moreover, RDFS' mechanisms permit to create classes of resources (that become instances of classes) sharing common properties as well as relationships among these resources.In their turn, classes are resources too, and any class may represent a subclass of another.This hierarchical semantic information structure is what allows computers to determine the semantics of resources based on their properties and classes.
5. The OWL 17 , built upon RFDS, is the richest standard description web vocabulary available today to define web ontologies used to create advanced Semantic Web applications (O'Connor et al., 2008).These ontologies consist of a taxonomy (system of classification) 18 and a set of inference rules from which automatic logical deductions (conclusions) can be made.The OWL syntax (e.g.subClassOf, disjointWith,  Iryna SOLODOVNIK   unionOf, intersectionOf) allows to assign properties to classes of resources as well as permits their subclasses to inherit the same properties.The growing expressive complexity of OWL finds its fittings in three sub languages, such as OWL Lite, OWL DL, and OWL Full19  (Lacy, 2005), each with enhancing level of detail required by different web semantic models.
6. Statements built on RDF structures are queried by means of SPARQL (Simple Protocol and RDF Query Language) 20 .
Here below (Figure 2) there is a graphical representation of the Semantic Web technologies.

Figure 2. The Semantic Web Technologies Stack 21
As we can see from Figure 2, over the ontological level of the Semantic Web stack, there is the Logic level managed by SWRL 22 language.This level should provide automated reasoning and inference of machine understandable knowledge, allowing it to be automatically integrated and reused by web applications.To achieve the full potential of the Semantic Web, information must be approved at the Proof level, permitting humans to retrace the steps a Semantic Web agent took to arrive at a particular conclusion.Finally, the entire Web would benefit of reliability and security (Weitzner et al., 2007) of web information validated through digital signatures at the Trust level23 .Digital signatures are encrypted blocks of data that computers and agents can use to verify that the attached information has been provided by a specific trusted source.As Miller (2009) put it in one podcast, "the Semantic Web will expose all of the problems of the Web like trust, provenance, and reliability (problems which are already very much with us) in a large distributed space".
One of the fundamental problems of the Semantic Web is to make available various types of data, so that they could be integrated and interoperable.Technically, it can be achieved through the appropriate technologies converting different datasets formats into RDF."The process of converting all existing data to RDF can be a major hurdle for organizations with large numbers of unstructured text documents and few metadata experts.Many tools have been developed to help automate named entity recognition, which is the process of using software to automatically identify and classify text elements like the names of persons, organizations, geographical locations, expressions of time, or expressions of quantity" (Goddard et al., 2009).
The available technologies such as, for example, POWDER 24 , RDFa 25 GRDDL 26 , R2RML 27 , RIF (Kifer, 2008), Drupal7 (Corlosquet et al., 2010) allow to make an automatic RDF mark-up of websites.To transform structural data in RDF/OWL formats there are tools such as Web services links & resources 28 ; SemWev29; Beckett30, SIMILE (RDF crosswalks) 31 , Semantic Bank (Huynh, et al., 2005), D2R Server (exposure of related databases in RDF) (Bizer et al., 2009), Altova SemanticWorks (the ground-breaking visual RDF/OWL editor) 32 .Moreover, to convert unstructured text into blocks of main entities, topics, reports; as well as to perform extraction of keywords, auto-tagging and disambiguation of entities and concepts -which may serve as outputs for RDF -there are several semantic tagging tools API (Application Programming Interface) such as, for example, OpenCalais 33 e Zemanta 34 .
Despite the existence of different tools for automatic RDF metadata publishing (Berrueta et al., 2008), included for semantic link discovery (Hassanzadeh, 2009; Volz et  al., 2009), the development of links35 between different datasets is not a trivial process, because it is necessary to carefully calculate an organic re-use of data shared within different user communities.It is also worth noting that the process of creating links may be carried out both through manual and ad hoc algorithms, expressing explicitly properties and values of (meta)data and constraints imposed on them.Anyway, considering how different technologies have been proliferating to support the creation of RDF/OWL, it is likely that the Semantic Web vision is one that will be realized globally in the near future.
"Exposing data as RDF is an important first step, but to actually achieve the linkeddata vision we must set explicit RDF links between data items within different data sources.This provides the means by which we can discover more information about a given entity" (Goddard et al., 2009).To actually link the Semantic Web datasets, in 2006 Tim Berners-Lee -in his memorable web document "Linked Data"36 -proposed a new Semantic Web technology called Linked Data (LD).LD is a technology based on: 1. RDF (to provide useful information on the object), 2. Hyper Text Transfer Protocol/HTTP (so that these objects can be referenced, searched and accessed by user agents), and 3. dereferenceable URIs identifying objects,."emphasizing data interconnections, interrelationships and context useful to both humans and machine agents"37 .
By means of these supporting tools, LD provides best practices for publishing (Bizer  et al., 2007), exposing, connecting and sharing different data(sets)38 across the web.The main purpose of LD is to break down the technological barriers that prevent free data sharing (Bizer et al., 2009), and to enable more powerful exploration of linked datasets (Alexander et al., 2009) structures through SPARQL queries (Cyganiak et al., 2008).However, according to some authors (Bechhofer,S., Ainsworth,J., Bhagat,J., Buchan,I.,  Couch,P.and Cruickshank, D., 2010), LD still misses a mechanism describing the aggregation of resources making their relations well interpretable, in order to capture better the added value of data collections and to allow its reuse through the exchange of a single object.
In Figure 3 there is provided a graphic representation of LD datasets39 published on the web.Referring to the practical use of LD datasets, imagine a system such as, for example, an Institutional repository (IR).Here many contributions (articles, books and their parts, thesis, conference proceedings) containing good bibliographical indications (including those of websites) may be deposited.The adoption of linking mechanisms could enrich and enhance these indications through connecting (Hennekenet al., 2011; Woutersen-Windhouwer, 2009) them with other citations, entries of encyclopedias, glossaries, classifications and other value (controlled, authority) vocabularies published on the web as LD structures.
The relative abundance of potential links can be, in their turn, enriched with thousands of other links pointing to another information resources, registers of agents (people) and their curricula, organizations, generating a powerful cross-border flow of information and data."Making entities identifiable and referenceable using URIs augmented by semantic, scientifically relevant annotations greatly facilitates access and retrieval for data which used to be hardly accessible" 44 .Really, LD publishing, sharing and interlinking scientific resources and data is intended to extend and fully realize the potential of access to scientific resources and collaboration within and across disciplines, whose knowledge is exposed and conveyed on the Web (Heath et al., 2011).
To control the quality of data exposed through LD mechanisms, there is a need to validate it by means of authority data inherent to Knowledge Organization Systems (KOSs) 45 .KOSs consist of authority systems such as thesauri 46 , classification schemes, subject heading lists, taxonomies and others controlled vocabularies.To port the already existing KOSs on the web (Tudhope , 2004; Zeng, 2009), as well as to provide conceptual modeling language for developing and sharing new KOSs, the W3C has developed the Simple Knowledge Organization System (SKOS).Particularly, SKOS is an application of the RDF 47 and its details have been released in "SKOS Reference" 48 together with a user guide "SKOS Primer" 49 .SKOS is aimed at building a bridge between KOSs (used in libraries, archives, museums, government portals, enterprises, social network applications and other communities) and LD community, bringing benefits to both.Indeed, nowadays, "Libraries, museums, newspapers, government portals, enterprises, social networking applications, and other communities that manage large collections of books, historical artifacts, news reports, business glossaries, blog entries, and other items can use SKOS to leverage the power of Linked Data" 50 .
SKOS represents an excellent way for conceptual exposure, management, sharing and re-use of authority data on the web, linking them with related authority data and integrating with different metadata conceptual schemes [e.g. with Dublin Core 51 , Library of Congress Subject Headings, MARCXML (Summers et al., 2008)].This is possible by identifying concepts provided by KOSs with URIs, labeling them with strings in one or more natural languages, as well as documenting them with different types of note, semantically related to each other, and aggregated into concept schemes.The growing scenario of use cases 52 implementing SKOS offers the prospect of linking together vocabularies provided by different sectors, thus enhancing "authority control" 53 of their data exposed on the web.The task to validate data through authority data ensures compatibility between different datasets, as well as their harmonized automatic management and interoperability at an aggregate level."Finally, the SKOS vocabulary itself can be extended to suit the needs of particular communities of practice or combined with other modeling vocabularies" 54 .
A knowledge global network sharing datasets outputs enhanced with LD and SKOS approaches is a perspective of a few years, considering also that there are enough good practices and use cases to be imitated 55 .

Linked Open Data and Supporting Experiences
The LOD cloud diagram, already presented, assumes that different datasets "must be provided in such a form that there are no technological obstacles to share data.This can be achieved by the provision of the work in an Open Data format, i.e. one whose specification is publicly and freely available and which places no restrictions monetary or otherwise upon its use" 56.This means, that the LD paradigm matches well with the vision of Open Data.
According to Open data definition, Open data, is a piece of open knowledge that is "free to use, reuse, and redistribute" 57 under an unrestricted license such as, for example, Creative Commons (CC) licenses, and Talis 58 (Campbell et al., 2010).The goals of the Open Data are inherent in several Open movements, such as Open access 59 , Open content 60 , Linked open science 61 , Open Knowledge 62 , Open Government 63 , Open Bibliographic Data 64 , Open Source (Buitenhuis et al., 2010) and others.
Open Data together with Linked Data 65 aim to break down the social, cultural, legal and economic barriers freeing shared data between human and software agents.Newly released Open Data published with LD mechanisms may be directly linked to the already existing open datasets (e.g.DBPedia.org,Wikipedia, WikiGuida, Geonames, MusicBrainz, lexical ontology WordNet, DBLP bibliography) exposed within the LOD cloud, thus reducing duplication of data and, above all, keeping the data updated 66 , and offering different agents to discover new information and to create and share freely new knowledge.
As the Open Data and LD paradigms are developing into a mainstream topic, more and more organizations are announcing new projects and services making their data open and publishing it as LD.Furthermore during the last two years Open Data (Danowski, 2010) and LOD has received a lot more attention from the library world (Dunsire, 2012).Here are some examples of related, practical experiences: ◊ The Harvard Library Policy on Open Metadata is committed to providing "Open Access to library metadata, subject to legal and privacy factors.In particular, the Library makes available its own catalog metadata under appropriate broad use licenses" 67 .
◊ The Library of Congress has developed the 'Authorities and Vocabularies' 68 service exposing its terminological systems and standards in an open manner and publishing them as LD.
◊ The German National Library (DNB) has developed 'Authority Data Linking' connecting its bibliographic data with Wikipedia, DBpedia, and VIAF datasets (Keßler,  2010).In cooperation with German Serials Database (Zeitschriftendatenbank -ZDB) 69 , the DNB has also generated Linked Data Service 70 .This service publishes DNB's bibliographic data and authority data as LOD, under the CC0 license, making such data available according to Open definition 71 .The record structures expressed in RDF/XML are available on the portal DNB 72 representing an experimental service which is going to be continually expanded and improved in accordance with the transparent procedures of the public domain 73 .
◊ The Hungarian National Library has published its bibliographic and authority data in open modality, using RDFDC, FOAF, and SKOS 74 .
◊ The British National Library (BNB) has provided its data via RDF download 75 .This practice of converting data is going beyond the encoding of collections of MARC records (over 2.8 million) in RDF/XML.Moreover, the BNB is working in the direction of making available the British National Bibliography (BNB) 76 4).
The EDM model, showed in Figure 4, enables to reuse the RDF, RDFS, OAI-ORE 88 , SKOS, DCMI Terms 89 namespaces (describing digital bibliographical records), outputting bibliographic data in LOD.Europeana's professional knowledge sharing platform 90 , based on this model, is a multi-lingual online collection of millions of digitized items derived from European museums, libraries, archives and multi-media collections.Europeana projects converts different terminologies and other KOSs (provided by various cultural institutions) 91 in SKOS, publishing them as LOD 92 .and their vote to support the Open licensing of their data to groups like LOD-LAM (Linked Open Data in Libraries, Archives, and Museums) 99 (Oomen et al., 2012), IFLA'S Semantic Web Special Interest Group (SWSIG) 100 , to library system vendors and providers, discussing and experimenting with LD technology, clearly reflect that LOD has become gained a lot of impetus in science, library and other cultural domains.The question is how to ensure that LOD won't be a temporary hype but that it will take hold in future infrastructures, generating LOD datasets from legacy systems and promoting the LOD approach towards a global and open information space.
To support the promotion of Open Data, different communities make their efforts to elaborate principles for releasing Open Data.Thus, the Open Bibliographic Data Working Group of the Open Knowledge Foundation has recently published "Principles for Open Bibliographic Data" 101 .
The concept 'bibliographic data' refers to data [e.g.author(s), title, publisher, date, title, page information, format of work] describing bibliographic resource as a unique resource in the set of all bibliographic resources, indicating also the modality how to find [e.g.URL address; URI identification: URN, DOI; ISBN, LCCN, OCLC number; links to related content, etc.] the described resource.
Formally, the Open Bibliographic Data Working Group recommends to release open bibliographic data or its sets with clear and explicit license 102 statements regarding re-use and re-purposing of bibliographic elements.The licenses such as Creative Commons licenses (apart from CC0), GFDL, GPL, BSD (with non-commercial and other restrictive clauses) are considered not appropriate to release Open Data, because they hinder to effectively integrate and re-purpose datasets, preventing also commercial activities that could be used to support data preservation.The "Principles" establish that open bibliographic data should be explicitly placed in the public domain via the use of the Public Domain Dedication and License (PDDL) or CC0 Developing recommendations, upcoming challenges, and technical alignment of catalogues and legacy systems in cultural institutions, and authoring environments for scholarly communication with open data and service infrastructure based on Semantic Web principles will be the strategic ones for practical promotion of LOD approaches.Moreover, the related experiences needs to be supported not only by "education" of interested parties regarding the use of correct licenses and LD techniques, but rather

Other Use Cases of Linking Data at the Semantic Level
Making explicit links among different data, especially those at the semantic level, requires careful analysis and rigorous definition of all necessary features of a (meta) data system.As it was already mentioned, this can be achieved through the definition of a formal explicit and shareable specification (ontology) identifying concepts, their properties, values and relationships defining granularity of knowledge of a reference domain (Valkeapää et al, 2007).To link data of different knowledge domains exposed on the web, it would be a good practice to establish a common ontology (Sure, 2005;  Vockner, 2011) for data sharing based on the already existing and widely used ontology structures 103 .Establishing a common ontology will make data interpretable in a shared manner, thus also helping to create highly interoperable semantic web applications and services 104 (Sanfilippo et al., 2003; D'Aquin et al., 2008).
To show how data could be linked by means of common ontologies, here will be worth mentioning a cross-institutional project called "ResearchSpace" 105 carried out by the British Museum.This project aims at harmonizing data provided by different cultural organizations, using RDF to set up mechanisms for the semantic search 106 .Particularly, the project uses a high level ontologyaimed to improve search accuracy by understanding searcher's intent and the contextual meaning of terms as they appear in the searchable dataspace.This ontology is based on the Conceptual Reference Model (CIDOC-CRM) 107 representing a data framework mapping links among different thesauri 108 terms supplied by "ResearchSpace" 109 users.The purpose of this methodology is to allow structured semantic search across multiple heritage repositories connected toGeoNames 110 exposed LOD cloud Joining the LD movement, in 2011 Google, Yahoo and Microsoft have agreed to adopt a common ontology maintained by Schema.org 111 .This ontology permits to publish linked structured data on the web, allowing different applications to create intelligent services systems (e.g.UMBEL Web Services, Virtuoso Universal Server, Linked Open Data Around-The-Clock) 112 .Recently, the already mentioned "Europeana" has announced its new project called Europeana Libraries 113 , which aims to integrate into one research area digital collections of the best digital libraries from 11 European countries.The thematically categorized collections will be linked to the Google Books 114 and to other web collections of photographs, manuscripts, historical films, to PhD thesis harvested by DART-Europe 115 , as well as to scholar articles from DOAJ (Directory of Open Access Journals) 116 .The Europeana Libraries collaborative platform will enrich data through a common ontology matching the ontologies of European libraries, and increasing the retrievability and re-use of their collections.In the light of the DRIVER (Digital Repository Infrastructure Vision for European Research) project, there was created an ontology-driven platform for semantic annotations to the 'Academic Institutional Repository and Bibliography' 117 .The project has led to the discovery of new ways of semantic data exchange and, consequently, to the improvement of harvesting of the semantically related data.In order to link different disciplinary fields of UK High Educational institutions, the JISC-SemTech Project 118 (carried out by the University of Southampton) developed several ontology-based applications (connected through LD) for semantic search.
Within the MIT Libraries Cataloging OASIS project it was implemented the openly available Utility Tool 119 converting MARC (Machine-Readable Cataloging Standards) 120 and MODS (Metadata Object Description Schema) 121 into RDF .RDF obtained data can serve as output for a subsequent modeling of common ontology based on RDF.Also The 'Bibliographic Ontology Specification' (D'Arcus et al., 2009) has provided the general concepts and properties useful to publish citations and bibliographical references (e.g.books, articles) through the Semantic Web ontologies, relying on CCL licenses and RDF technologies.
The 'Bibliographic Ontology' (Bibo) 122 provides a good practical example of publishing bibliographic data in RDF.The Bibo may convert in RDF a wide range of current metadata formats.Another experience is an Open catalogue of the world's cultural works (books, music, films) called 'Bibliografica" 123 , which runs on the OpenBiblio software 124 .'Bibliographica' offers an ontology-driven platform based on the native RDF linked data support, FRBR-like domain model, and Wiki-like recording of every change of bibliographic data.'Bibliographica' allows to create personalized collections, to add additional information to bibliographic entries, and to share these last with Wikipedia.
In this context, it is also worth mentioning the 'MarcOnt' 125 , which through the integrated RDF Translator provides a technique for common ontology integration of bibliographic descriptive formats such as MARC21, BibTeX and Dublin Core.Meanwhile, the released MODS Ontology 126 represents a good ontology strategy for the migration of MARC metadata into MODS 127 expressing its entries in RDF and OWL.This make bibliographic data acceptable by the producers of LD.
Hereafter, within the UCSD Libraries' Digital Library Program 128 , that discovered some limits of the DSpace 129 and Fedora 130 software regarding the acceptance of some bibliographic data formats, there was developed the ARK [Archival Resource Key] tool (Kunze, 2003).This tool allows to transform hundreds thousands of MARC and MODS data in RDF, further loading it into the AllegroGraph RDF 131 queried by the SPARQL language.Beyond this experience, new versions of Eprints 132 and DSpace (Bosman,  2009) software were released allowing to publish bibliographic data in RDF formats, to customize and qualify this data through semantic Dublin Core Application Profiles such as SWAP, IAP and TBMAP, and to connect different data through LD mechanisms.Iryna SOLODOVNIK In this new, and certainly not exhaustive, scenario of semantic exposure of bibliographic data on the web, it is important that more and more information providers (included cultural, scientific and administrative bodies) make their data available in formats adaptable to the Semantic Web, replicating current experiences and proposing new projects, tools and use cases.

LODe-BD: Enabling Bibliographic Data to become Linked Open Data
The concept 'LODe' refers itself to the concept of 'Bibliographic Data' (BD), forming together a complex concept such as 'LODe-BD' .Some authors 133 recognize for the 'LODe' , particularly, for the final 'e' something that can be 'e'-mbedded within the system itself.On the other hand, the AIMS (Agricultural Information Management Standards) 134 (Subirats, Nicolai and Waltham, 2010), team defines 'LODe' as LOD-'e'-nabled, where "enabled" is the potentiality of data to become Open and Linked (LOD) data.
To publish BD as LOD data, there must be identified standards, formats and licenses able to support BD within the LOD cloud space.There should be a common agreement on the data exposure as well as "minimal set of properties meaningful in data sharing" (Subirats, and Zeng, 2011) in the LOD data space.To assist this task, the AIMS team has developed and posted on its website the Recommendations LODe-BD 135 (Figure 5).These Recommendations provide the necessary steps and assessment tools to support agents in choosing strategies and standards for encoding BD as LOD.Particularly, the Recommendations, based on five key principles, provide a set of instructions and tips enabling structured bibliographic data describing digital resources (such as articles, monographs, thesis, conference papers, presentation material, research reports, learning objects) 136 to acquire LOD characteristics.The five key LODe-BD principles call: 1.To promote the use of well-established metadata standards as well as the emerging LOD-enabled vocabularies proposed in the Linked Data community; 2. To encourage the use of authority data, controlled vocabularies, and syntax encoding standards whenever possible in order to enhance the quality of the interoperability and effectiveness of information exchange; 3. To encourage the use of resource URIs as names for things for data values when they are available; 4. To facilitate the decision-making process regarding data encoding for the purpose of exchange and reuse; 5. To provide a reference support that is open for suggestions of new properties and metadata terms according to the needs of the Linked Data community.
The LODe-BD Recommendations not only provide information on how to publish and use open bibliographic data as LD, but also on where to retrieve LD sets and vocabularies supporting the LD publishing 137 .Iryna SOLODOVNIK As a first step enabling BD to move towards LOD, the Recommendations offer the descriptive guide about necessary properties of bibliographic metadata, arranging them in nine groups (Table I).This group contains the properties associated with the Agent (Сreator, Contributor, Publisher) responsible for creating and publishing of the resource content.

Physical characteristics
Properties that describe the appearance and characteristics of the physical form of a resource.They are: date, identifier, language, format, edition / version.

Collocation
It is considered important for a resource to be located and retrieved in the area of information exchange.The properties of this group are represented by location and availability of a resource.

Subject
In contrast to the physical characteristics, the group «Subject» encompasses properties describing or, at least, helping to identify what a resource denotes (under the subject term, classes/categories, keywords assigned geographic entity).

Description of the Сontent
There are two main types of descriptions being focused on the content of a resource, rather than on the physical object: a) every description is representative of the content, usually in the form of abstract, summary, notes, table of contents; b) the type or kind of resource.

Intellectual property
Any property that is concerned with the intellectual property rights relating to access and use of a resource, with particular regard to the rights, use and access conditions.

Use
The properties that relate to the use of a resource, rather than the characteristics of the resource itself.Typical characteristics are: users, their level of education.

9.
Relationship between documents/ agents (responsible for the creation / publication of documents) This group defines the relations/connections between two resources or between two agents.Considering the significant number of properties of the connections, the specific properties of the relations are explained in other parts of the Recommendations.
Source: <http://aims.fao.org/lode/bd/properties> The nine groups are further extended with the specific properties, presented and explained in Table II.

LODe-BD Decision Trees
The decision trees of LODe-BD are designed to assist any bibliographic data provider in the metadata selection process.Particularly, these decision trees are based on flow charts that guide the choice of properties included in the already mentioned nine groups of metadata.Moving from the property describing a resource instance, each diagram shows the flowchart for a decision point, offering a progressive solution for encoding metadata (by means of symbols presented together with their description in Figure 6).  it is important to provide passages useful for data developers to understand how to manage the available data.These passages are graphically explicated through the Concept Maps and flowcharts.In Figure 7 the decision trees and the explanation how to encode the properties of "Title" and "Creator" are reported.Each decision tree provided for each property and explained in its details is designed to facilitate the selection of the appropriate strategies to design semantic data models validated by means of authority control and based on the standards proper to the communities involved in bibliographic data management.Moreover, such a design should take into consideration the concept of total openness (Open Data) and accessibility of LOD-e BD.
Once it is decided to publish a bibliographic database in the LOD modality, there should be decided what types of entities and relations are involved in the description of bibliographic resources.For this purpose, a LODe-BD concept model is introduced (Figure 8).It aims at sharing common understanding on creating entities and relations of semantically rich BD.The LODe-BD concept model is developed on the basis of the FRBR (Functional Requirements for Bibliographic Records) (Saur, 1998), by means of which it is possible to extend and reconsider significantly the LODe-BD strategies in modeling semantic metadata.In the left part of Figure 8 there is provided a high level abstraction of LODe-BD general concept model describing the central entity of any information system -Resource, related to Theme and Agent.In the right part of Figure 8, the implication of the general concept model in the LODe-BD is shown and the examples of possible relationships among instances of different entities are provided: 1.The entity Resource is the starting point of any bibliographic description in LODe-BD decision trees.
2. Relationships are established among the entity Resource and other two major entities: Agent: the entity responsible for creating of the content and/or for dissemination of the resource; and Theme (subjects, themes/topics, concepts and categories of the created content).
3. There may be also created relationships among instances of an Entity.For example, a Resource can be connected to another Resource.An Agent may be related to another Agent.
4. The relations between any pair of instances vary and may be created at different levels.For example, an Agent can provide funds for the creation of an original work, for the translation of this work, or for the release of a new format of the translation.
5. Authority Control (name authorities, value vocabularies) is considered an important element of the model.Agents, regardless of their role in the relation to a Resource, should be managed through authority files of names.In the same manner, through appropriate value vocabularies, title, main concepts (themes/topics), and geographical locations of the Resource should be controlled.Different authority files are already available in the LOD cloud.
The LODe-BD concept model represents one of the best practices enabling Bibliographic Data to get ready as LOD data.This model can also be used to mark internal, external and collaborative responsibilities of a LOD-enabled project highlighting each of its phases.

Standards for Metadata LOD-Ready
The Recommendations LODE-BD list widely-used metadata standards and emerging LOD-enabled vocabularies, that should be used to set up high-quality "LOD-ready metadata" (Table III) 138 .Despite the fact that the standards selected in the LODe-BD are focused on the knowledge domain supporting the Agriculture sector (AGMES), any other community modeling its knowledge datasets can adopt the LODe-BD as a reference model.It is only a matter of fact to select another list of standards appropriately.

Table III. Metadata Standards and Emerging LOD-Enabled Vocabularies
Source: <http://aims.fao.org/lode/bd/metadata-standards> The selection of the appropriate standards to get different metadata LOD-ready should fall on the choice of standards widely used in the reference community, as well as on the LODe vocabularies becoming increasingly popular within the same community.To guide the choice of right standards, the role of the "Decision Trees" approach remains specifically important providing assistance in the selection process presented through flowcharts and identifying the relevant properties in each of the nine groups 139 of metadata for LODe-BD.

Library Linked Data: From Collections to Connections
The continuous efforts of the W3C LLD XG Library Linked data incubator group 140 in exploration and creation of harmonized theoretical and practical issues guiding the implementation of connected bibliographical data and their collections on the web of data (Hanneman, 2010; Karen, 2010), have been resulted into development of recommendations concerning the publishing of Library legacy data as Linked Data 141 .Particularly, the "Library Linked Data Incubator Group Final Report" 142 , together with the supporting documents "Report on Use Cases" 143 and "Report on Datasets, Value Vocabularies, and Metadata Element Sets" 144 , released by the LLD XG in 2011, can be seen as a 'meeting place' among documented ideas of different professional communities 145 on publishing open bibliographic data as linked data under CKAN MetaData Conventions 146 .The graphical version of various datasets connected in the space of Library Linked Data (LLD) (Figure 9) is developed by the CKAN LLD group 147 .Although this graphical vision has still not been effectively 'translated' into a real Linked Data service, the conceptualization of different aspects of LLD represents a real effective start point to collect 148 and to program bibliographic data to enter into the LOD Cloud.
The colored circles of the graph represent the data packets developed within the domain LLD, while the gray circles represent the related data packets, most of which are coming from the LOD Cloud space (e.g.DBpedia, GeoNames, VIAF, DDC, LCSH, RDA, FOAF, DCMI Metadata Terms, the elements of RDA 149 ).Size of the circles and thickness of the lines reflect the size of the packets and the number of external links provided 150 .The dataset LLD may become the largest providers of authoritative datasets (e.g.names, subjects) 151 in the space of the Semantic Web as well as become easily enriched (authority data enrichment) 152 through authority records published as LD by other communities operating on the same Web 153 .
With reference to the digital environment, it is clearly important to control and to validate the wide range of digital data and their collections with the tools of authority control, and even more those ones to which libraries provide digital-only access.In particular, authority control will cover those links responsible to provide the description to subjects, individuals and corporate bodies and the identification of the role that they play in the creation, production, dissemination and preservation of different types of records.Iryna SOLODOVNIK Furthermore, by integrating into the Semantic Web, bibliographic datasets could interact widely with various web ontologies (Bai at al., 2011; Westrum et al., 2012) (e.g.SKOS, Library of Congress Name Authority File, BIBO, Lexvo.orgontology, CC REL, OPM, UMBEL vocabulary, WEMI FRBR ontology, ontology bibliography NSTL 154 ) drawing automatically from them the descriptions to enrich semantically their properties and connections.
Generating rich clusters containing bibliographic data and their collections, as well as their descriptive and normalizing properties, connected through the LD mechanisms, it will be of course also possible to process a variety of sophisticated queries on those clusters.Moreover, semantics of data connected by LD makes it easier to take advantage of a service that only partially matches a query request, involving the creation of a «value chain» in which sub-assemblies of information are passed from one agent to another, each one «adding value,» to construct the final product requested by the user query.
Despite the obvious benefits of LD 155 , the numerous projects 156 promoting the creation of innovative semantic services (Nandzik et al., 2010), the communities aiming at publishing their data on the web as LD should follow an accurate specific and decisive coordination, in order to counter the risk of irregularly implementation of these technologies.
Among the main obstacles to publication of data with the LD mechanisms, there is still reluctance on the different organizations to adopt new software platforms, as well as the lack of staff with adequate knowledge and skills to use these platforms before and after their effective implementation. 157 professionally trained librarians, as well as experts developing and managing metadata, may find themselves feeling overwhelmed by the prospect of producing semantic metadata and maintaining ontologies, other communities operating on the web may find themselves even more frustrated to support the similar tasks.
There are also tensions between open standards 158 , and open publishing159 required for the Semantic Web and existing methods supporting and maintaining the of data management infrastructures.From a more positive perspective, different web communities could certainly imitate the already existing good practices and use cases for publishing their bibliographic data with LD technologies.
As to the conclusions of the LLD Report, they are the following: ◊ bibliographic data are still not integrated with web resources; ◊ library standards have been designed only for the library community; ◊ bibliographic data are expressed primarily in natural textual language; ◊ the library community and the Semantic Web use different terminology for similar concepts of metadata; ◊ technological changes concerning library depend on the development of the largest vendors of digital information systems.
To counter the difficulties of integration of bibliographic data in the Semantic Web, the Report provides four main recommendations, with which: 1. invites library communities to identify their data packets (including thesauri and controlled vocabularies) (Vatant, 2010) to be published in LD mode and through Semantic Web standard languages (e.g.SKOS) in order to prepare them for a successful integration in the Semantic Web space, as well for creating of new and potentially unlimited services.Moreover, the Report invites libraries to spread their know-how on freeing data (Coyle, 2010; Danowski , 2010) 160 and in implementing of widely-used licenses161 allowing libraries to store, share, enrich and re-use data openly as needed.The Report identifies three potential categories of bibliographic data to be published in LD and to be managed through appropriate namespace 162 .They are: a. dataset forming the triples describing bibliographic resources, people, organizations and related entities; b. controlled (value) vocabularies from which there can derive the triples formed by values, properties and attributes for authority files; c. formats and vocabularies (schemas) of metadata generating the triples describing attributes and relationships within the bibliographic datasets and controlled vocabularies.
4. The report calls upon the organizations 163 producing standards for libraries.
Particularly, it invites them to increase the participation of libraries in the standardization of the Semantic Web (Dunsire, 2010), and in developing standards for bibliographic data compatible with LD.Moreover, it invites the same organizations to influence vendors (commercial companies providing digital resources and online databases) to consider and adopt these standards.
5. The report is addressing to developers of digital information systems to project high-level data management services based on the capabilities of the Semantic Web and LD.It also encourages to create URIs (even experimentally) for the elements of bibliographic datasets; to develop policies to manage effectively RDF vocabularies and related URIs; to express bibliographic datasets in terms of their reuse or mapping in connection to existing vocabularies published in LD.
6. Invite the communities of librarians and archivists: to preserve their datasets and controlled vocabularies in LD elements; to apply their know-how -referring to data management and long-term preservation -to datasets published in LD, so that LD triples become immutable, unique identifiers URI remain permanent, and names, titles, subject headings -as artifacts of cultural heritage -remain stable over time.
Observing the progress of the Semantic Web technologies, regional and national libraries together with archival institutions have a good opportunity to enrich their existing traditional roles of managers of the bibliographic universe with a new quality of the competition authorities 164 in the long-term preservation of data sets based on LD, in relation to cultural heritage.

Conclusions
The first step towards a web, that contains bibliographic data whose semantics is interpretable by machines, is surely to adopt an open standard format such as RDF that can be used to collect and categorize information scattered in the digital environment.The RDF will provide the data in their purest form (Raw Data 165 : The data free from formats) and will allow it to be structured and connected (through LD) in a homogeneous way within the space of the Web of Data.
A financial investment and that in know-how of experts from different professional areas, concerning digital data generation and management, will surely improve the quality of semantic technologies and architectures of LD.It will also certainly generate 163 IFLA (Namespace Technical Group reposting to Committe on Standards); JSC for Development of RDA (DCMI/ RDA Task Group, Dublin Core Metadata Initiative); DCMI Bibliographic Metadata Task Group (DCMI Vocabulary Management Community) 164 "Libraries and librarians have a very high "trust" factor with the general public, and we are in a position to raise the general quality of the Semantic Web by ensuring Library data competing with data from the widest range of sources", Dunsire, cit., 2012, http://www.nb.admin.ch/aktuelles/ausstellungen_und_veranstaltungen/00726/01611/03953/03958/index.html?lang=en 165 Berners-Lee.T., Talk on TED 2009, <http://www.ted.com/pages/about> a positive return -to harmonization of different formats of web-based data, to their reuse by different information systems and web communities, and to creation of new and powerful search tools for information retrieval through targeted semantic queries -over the medium to long term.
The challenge for managers of the bibliographic universe is to deepen the knowledge of LD, to know all benefits that LD could have for libraries, to promote the knowledge of LODe-BD and to advance in the Library Linked Data implementation.Currently, these themes represent a growing process aiming at making different communities more efficient and successful in aligning their data models and policies with the Web of data.The list of scenarios that could potentially benefit from the Semantic Web technologies and LD, as they continue to evolve, is limited only by the imagination.Considering the library community as a whole in a Semantic Web scenario, one authentic concern to it would be this one: "The technology is finally ready; it's critical for libraries to begin preparations to become full participants in the world of Linked Data" (Byrne et al.,  2010), "the first practical expression of the Semantic Web, useful and doable today, and applicable to all forms of data". 166

Figure 3 .
Figure 3. Connecting Web Dataset Through Linked Data Source: "Linking Open Data cloud diagram", <http://richard.cyganiak.de/2007/10/lod/>(clicking the original image will take you to an image map, where each dataset is a hyperlink to its homepage) The graph showed on Figure 3 is a result of efforts of the Linking Open Data community project 40 , within the W3C SWEO 41 group.The project uses the categories of datasets converging to a directory of Open Data datasets and Linked Open Data (LOD) called CKAN 42 managed by the Open Knowledge Foundation 43 .While in October 2007 the datasets of LOD cloud diagram were based on more than 2 billion RDF triples connected by more than 2 million RDF links (Berners-Lee, 2009); in 2011 the datasets counted 31 billion RDF triples connected by around 504 million RDF links.To determine whether LD technologies are sufficiently mature, there can be explored development and deployment of exposing data as RDF, and linking RDF entities together.

Figure 4 .
Figure 4.The Namespaces Used in the Europeana Data Model (EDM) Model Other experiences and initiatives ranging from: ◊ The knowledge-sharing platform "LinkedScience.org:Interconnecting scientific assets" 94 ; ◊ The Library of Congress' initiative "A Bibliographic Framework for the Digital Age" 95 ; ◊ The recently developed Bavaria's Open data portal 96 ; ◊ Sweden's national library system LIBRIS 97 publishing the Swedish National Bibliography along with the authority data under an open CC0 license provided as a complement to their LD implementation;

Figure 6 .
Figure 6.Symbols and Their Definitions in the Flowcharts LODe-BD Once having the tools and their descriptions in the Figures and Tables presented,it is important to provide passages useful for data developers to understand how to manage the available data.These passages are graphically explicated through the Concept Maps and flowcharts.In Figure7the decision trees and the explanation how to encode the properties of "Title" and "Creator" are reported. ,

Table I .
Groups of Common Properties (LODe-BD) for Bibliographic Data