Recent surveys, about the Linked Data initiatives in library organizations, report the experimental nature of related projects and the diculty in re-using data to provide improvements of library services. This paper presents an approach for managing data and its tacit organizational knowledge, as the originating data context, improving the interpretation of data meaning. By analyzing a Digital Libray system, we prototyped a method for turning data management into a semantic data management, where local system knowledge is managed as a data, and natively foreseen as a Linked Data. Semantic data management aims to curates the correct consumers' understanding of Linked Datasets, driving to a proper re-use.
data management of the local system relies on a tacit [Pol66] organizational knowledge that should be made explicit, as an ontology, in order to support the consumers’ understanding of the meaning of data. In this paper we address the problem of managing data and its knowledge for providing quality LD, driving data management toward the semantic data management.
Transforming data management practices into semantic data management practices requires to establish, in the holding ORG, a mindset specically oriented toward the pillar elements of the LD, like the Uniform Resource Identiers (URIs), and the Vocabularies (controlled term lists, thesauri, ontologies, etc.). This transformation implies to consider InfSyss’ knowledge as a data in the data management practices.
Semantic data management curates the correct consumers’ understanding and the correct interpretation of exhibited LD.
In this paper, we present two local ontologies obtained by analyzing the DigLib system and detecting the underlying tacit knowledge. We show how we have dealt with tacit knowledge capture, also in relation to existing ontologies, from dierent knowledge domains.
The matching between local ontologies and existing ontologies drives the production of LD datasets, whose meaning is supported also by the organizational knowledge , as it is mentioned by the LD best practices, and it is considered essential for the correct interpretation of LD. The remainder of this paper is structured as follows. Section 2 reports the state of the art of LD implementation in ORGs managing DigLibs. Section 3 provides an explanation of the semantic data management. Section 4 explains the knowledge types. Section 5 overviews the explicit knowledge of the DigLib system case study. Section 6 describes the method for classifying the tacit knowledge of the DigLib system. Section 7 presents resulting local ontologies as a LOV. Section 8 draws the conclusions and presents the future developments. 2
The 2nd Survey Report of the On-line Computer Library Center (OCLC) 1 published by Smith-Yoshimura in 2016 [SY16] reports the analysis of 112 Linked Data projects or services undertaken by 90 institutions in 20 dierent countries. The analysis, performed in 2015, describes the respondents’ motivations as publishers or consumers of LD and indicates that most of the initiatives are primarily experimental in nature. Among the most mentioned motivations we report the most relevant to this paper: (5) the need to publish linked data to consume it and to re-use it in future projects; (6) maximize interoperability and reusability of the data; (7) provide stable, integrated, normalized data on research activities across the institution. These motivations highlight the specic interest of disseminating and re-using data, that implies to consider not only the perspective of the end-user as a consumer, but also the perspective of the ORG as a data manager and provider that exposes LD.
The recent survey of Tosaka and Park [TP18] still report the signicant problem of the absence of comprehensive data, that could be used to guide improvements in continuing education for the library community. The survey identies specic knowledge gaps to be addressed by data repository systems and specically in relation to SemWebTech. The survey still reaches the conclusion of the exploratory stage of the LD implementation. 3
The management of data, and in turn the management of information and knowledge is one of the most studied and developed eld in the automation of information, which nowadays is challenged by the automation of Semantics conveyed by the hierarchy of data, information, knowledge and wisdom [Row07]. The main dierence between data management and semantic data management is that, in data management the semantic context of data is managed by persons aliated to the ORG, by means of their human information, implicit and explicit knowledge, wisdom [Row07], while in the semantic data management, the data is equipped with its semantic context, which is codied in a machine-interpretable form (SemWebTech). The semantic data management provides machines with data and its knowledge context in a SemWebTech form, and the interpretation can benet both machines and humans. Machines might re-use or re-manage data in a proper way, supported by knowledge driving the understanding of why data has value, humans might be unloaded by long and discontinue searches of additional information for understanding of why data has value. Nevertheless capturing relevant knowledge context for data is challenging. The literature is rich of work generating ontologies from data and metadata of relational databases, but the underlying knowledge not always 1http://www.oclc.org/research.html is explicit and is scattered into software documentation (also not always well-documented) or into technical reports, and its retrieval often is a time-consuming task.
The tacit knowledge, inadvertently implicit, hidden and given into the data management practices, is essential for enabling machines to properly interpret data, thus its detection, and its codication as a LOV, is an essential part of the semantic data management.
The adoption of SemWebTech in an existing data management system, implies indeed to understand how data can be interpreted by a third party, and as such, how the explicit and tacit knowledge about data management systems, has to be captured and provided as the data context. 4
The ORG knowledge is a research eld developed since early nineties [Wii94]. Taking into account the most cited work of the eld, comprehensively described by Evans et al. [EDB15], we review some theoretical analysis of the eld supporting the method experimented for managing and structuring the ORG knowledge. The ORG knowledge, as the data context, should be comprehensive of explicit and tacit [Pol66] [Gra96] knowledge.
The explicit knowledge in the knowledge management literature is distinguished in:
Codied knowledge or knowledge that can be stored or put down in writing without incurring irreparable losses of information [Cho96]. This form of knowledge is highly rened [Wii94] and formalized, which allows it to be disseminated, more easily, more rapidly, and more extensively in the ORG than other forms. Encapsulated knowledge, which is not fully codied, and it is object-based, since the substantive knowledge that went into the design and development of artifacts remains partially hidden from its users [vdB13]. This is exactly the case of the RDB, where substantial expertise has been spent in its design as well as in its data population.
Encapsulation consists of the transformation of substantive knowledge into a product that requires only functional knowledge for its utility [vdB13]. Extracting and codifying encapsulated forms of knowledge requires further unpacking using methods similar to reverse engineering or compositional analysis.
However, encapsulated knowledge is dicult to be collected and may also be subject to a misappropriation. The tacit knowledge denitely is the source of the codied and encapsulated [vdB13] knowledge, which provides the grounding of meaning and the basis for the interpretation of a tacit activity [EDB15]. 5
The Explicit Knowledge of a Digital Library System 2Metadata Encoding Transmission Schema, www.loc.gov/standards/mods/ the PREservation Metadata Implementation Strategies (PREMIS) [PRE15] for managing preservation metadata about multimedia objects; the Metadata Encoding and Transmission Standard (METS) 3 for DigRes packaging.
For the purpose of this article we focus on MODS and PREMIS, as a sample of explicit knowledge managed by the MassConv system. By analyzing the data content, of such explicit knowledge, we observe that data can be used by other relevant ontologies from other knowledge domain, and the exhibition of these knowledge connections as LOV allows data to be interpreted also by consumers, coming from other knowledge domain. Thus in the following subsection we briey present the ontologies strictly representing the MassConv explicit knowledge (MODS and PREMIS) and the ontologies extending the interpretability of data, the Organization Ontology (ORG-O) and the Provenance Ontology (PROV-O). 5.1
The Codied Knowledge Matching with Dierent Domains The explicit knowledge about data, managed by the local system, is formally codied in the following LOVs [VAPVV17]:
Metadata Object Description Ontology (MODS-O) MODS-O develops around the main class mods:ModsResource which represents any library-related resource such as a book, journal article, photograph, or born-digital image that is described by a MODS resource description.
PREMIS Ontology PREMIS-OWL models the knowledge domain of digital preservation metadata, and develops around four main classes: premis:Object, premis:Event, premis:Agent and premis:Rights.
Provenance Ontology (PROV-O) PROV-O [W3C13b] describes the concepts related to the provenance in heterogeneous environments, and develops around three main classes: Agent, Entity, and Activity. Organization Ontology (ORG-O) ORG-O [W3C13a] develops around the core class org:Organization which represents a collection of people organized together into a community or other social, commercial or political structure.[..].
3Metadata Encoding Transmission Standard, www.loc.gov/standards/mets/
Detecting Tacit Knowledge in the Local System By analyzing the MassConv database, we have realized that the explicit knowledge stored into the RDBs (data and schema) contains part of the system knowledge, that has driven its development. Commonly, the timeliness and costs plays the main role for reaching observable results. This fact drives worker stas to neglect the capture of knowledge, produced during the period of system development. During the MassConv development, most of the knowledge locally created and used (see de Vasconcelos [dVKCR17] mapping between Knowledge Life Cycle and Software Development Life Cycle), remains scattered in text documentation which is dicult to be retrieved, and to be systematized, thus the knowledge context of data cannot support the understandability of data. This problem is unavoidably inherited by processes dealing with the generation of LD from the MassConv RDB.
Thus we have adopted the method of a) collecting software functions parameters passed through the massive conversion workow; b) matching written denitions in the text documentation; c) creating identiers for parameters as piece of embedded knowledge; d) creating a local ontology as the knowledge artifact for turning tacit>embedded knowledge into explicit knowledge; e) matching local ontology to existing ones. Consequently, tacit knowledge was captured and formalized in a local ontology, expressing the knowledge underlying the MassConv system. 6.1
Ontology for Software Embedded Knowledge The implicit founding concepts for the management of SDL InfSys digital assets, were codied into a local ontology, named On-SDL. The main classes of the ontology are:
The Organizational Collection provides an abstraction layer documenting the evolving history of the physical ORGs (premis:Agent), dealing with dierent legal aspects ( premis:Right), and its changes (premis:Event) involving the maintenance of the Digital Objects ( premis:Object).
The Digital Metadata Object is a text le of whatever format (XML, CSV, JSON, RDF) comprehending data and metadata describing premis:Objects managed by the DigLib system.
The Digital Content Object is whatever resource, (a multimedia le, a database...) which needs to be managed by the SDL system.
These classes are the parameters most used by the software functions, composing the MassConv system. The ontology formalization for representing the main concepts, roles and individuals, of On-SDL, have been initially expressed in ALC the basic DescLogs [NB +03].
A DescLogs Ontology O consists of a TBox T , and an Abox A, respectively representing the intentional and the extensional knowledge [Baa03].
Figure 2 depicts the TBox T modelling the intentional knowledge, managed by the MassConv software: UniversityORG subClassOf UniversityDL : Sapienza organizational structure is reproduced in the SDL. UniversityORG subClassOf prov:Agent: Sapienza ORG are type of PROV-O agents.
org:Organization equivalent UniversityORG : Sapienza is a type of ORG. premis:Object equivalent DigitalObject : PREMIS Object is equivalent to DigObj.
DigitalCollection subClassOf DigitalResource : DigColl is a type of DigRes.
The ALC roles are expressed in domains and ranges, by using the existential quantier quantier 8: 9 and the universal UniversityORG manages DigitalResource : ORG manages at least one individual and all those individuals are Digress.
UniversityDL aggregates DigitalResource : DigLib aggregates at least one individual and all those individuals are DigRess.
DigitalCollection collects DigitalResource : DigColl collects at least one individual and all those individuals are DigRess.
DigitalResource contains DigitalObject : DigRes contains at least one individual and all those individuals are DigObjs. MassConv was developed for managing the DigRes production workow, based on Information integration GlobalAs-View approach [Len02], where the RDB is the data management technology (the data source S), that is mapped by M, toward global schema or ontology G. Figure 3 depicts on the right this process and on the left, the MassConv workow steps, that are described as follows: 1. Organizational Collection creation identies (with a root URI) a Sapienza ORG. 2. Object Acquisition stores DCOs from a Sapienza ORG, into a working area, assigns to the DigRess, the URI based on the ORG identier, associates the related descriptive data (MODS), and computes or collects preservation data (PREMIS). 3. Mapping Development builds the conversion layer toward MODS and PREMIS semantics. 4. Object Accessioning stores DCOs in the SDL repository, from the Acquisition working area, propagates and extends DigRess’ URIs over belonging DCOs. 5. Collecting Preservation Metadata about DCOs is automatically gathered and computed. 6. Digital Resource production DMOs and related DCOs are produced, according to the SDL XML metadata schemas G. 7. Linked Data production , the last step to be developed, re-uses data at the data source S, and extends the mapping M necessary to local ontologies G.
We can observe that each workow step is knowledge, embedded in the software functions performing each step. Consequently, that knowledge was codied into a local ontology for describing the MassConv Workow (MCWO).
Figure 3 shows on the right, how the tacit knowledge, that we have codied, ows from the data source S toward the On-SDL and the MCW-O, and in turn routes data toward existing LOVs, the PREMIS-OWL, the MODS-O, the PROV-O, the ORG-O.
Tacit Knowledge Codied as a Linked Data Vocabulary The method adopted for capturing tacit knowledge from the DigLib system case study is a prototype, that should be evaluated in other InfSyss. The detection of the tacit knowledge is in any case not a straightforward task, mostly is manual and requires the eort of the knowledge worker for being performed.
Nevertheless, the method can be a training for developing a mindset of knowledge workers, oriented toward the management of LD pillar elements, the URI and LOV. The transition into semantic data management, is based on the management of the system knowledge, likewise of the system data: knowledge data is codied in a machine-interpretable form. The computable interpretation of data allows machines to support better the human work in understanding of why data has value, and thus in re-using or re-managing data in a proper way. In the near future, we will be developing the workow for generating the foreseen Linked Dataset, referring to the described On-SDL and MCW-O ontologies. M , O G R O , O V O R P , W O S I M E R P o t d e t a l e r , O W C O L D S s i e
D-Lib [TP18] Yuji Tosaka and Jung-ran Park. Continuing education in new standards and technologies for the organization of data and information. Library Resources & Technical Services , 62(1):415, 2018.