A Computational Approach to Historical Ontologies
Mat Kelly, Jane Greenberg, Christopher B. Rauch, Sam Grabus, Joan P. Boone, John A. Kunze, Peter Melville Logan
AA Computational Approach to Historical Ontologies
Mat Kelly ∗ , Jane Greenberg ∗ , Christopher B. Rauch ∗ , Sam Grabus ∗ , Joan P. Boone ∗ ,John A. Kunze † , and Peter Melville Logan ‡∗ College of Computing and InformaticsDrexel UniversityPhiladelphia, PA, USA { mkelly, jg3243, cr625, smg383,jpb357 } @drexel.edu † California Digital LibraryUniversity of CaliforniaOakland, CA, [email protected] ‡ Temple LibrariesTemple UniversityPhiladelphia, PA, [email protected]
Abstract —This paper presents a use case exploring the appli-cation of the Archival Resource Key (ARK) persistent identifierfor promoting and maintaining ontologies. In particular we lookat improving computation with an in-house ontology server in thecontext of temporally aligned vocabularies. This effort demon-strates the utility of ARKs in preparing historical ontologies forcomputational archival science.
Index Terms —Archival computation, archival science, histori-cal ontologies
I. I
NTRODUCTION
Over the last 30 years, governmental research agencies,foundations, and institutions have collectively allocated mil-lions of dollars toward digitization of nearly every type ofarchival artifact, regardless of genre or format. A necessarypart of these processes includes the creation and integrationof metadata to support resource discovery, access, and use.As these collections amount to big data, there is increasedattention toward computational approaches with metadatain the context of computational archival science [1], [2].Computation-ready metadata, increasingly referred to as smartand big metadata, can help contextualize an archival collectionand offer new insights into the resources being described [3],[4]. Unfortunately, the full exploitation of topical or subject-oriented metadata is hampered by the absence of temporallyrelevant ontologies and the challenges with persistence amongontological resources terminologies.Large scale digital archival collections simply use contem-porary ontologies such as the Library of Congress SubjectHeadings (LCSH), one of the most expansive, general domainterminologies available in linked data. A key limitation is thatcontemporary ontologies, despite being computation-ready, areat odds with providing contextual interpretation and analysisof historical archival resources. Researchers may use wordembeddings and data mining to understand historical conceptsand the context of a particular collection [5]; however, thisapproach is not integrated with ontology standards and lacksthe semantic infrastructure that can help support the study ofconcept and context over time. Research applying ontologiesto historical records requires a more holistic approach, one thatcan be embedded in the practice of computational archivalresearch. Specifically, there is a need for computation-readyontologies that temporally align with the resources being represented. Research in this area also needs to consider per-sistent identifiers (long-term, stable data links) for ontologicalconcepts to support longitudinal analysis of digital archivalartifacts over time.
Fig. 1. A catalog entry for the LCSH1910 vocabulary.
Initial work in this area has been pursued as part of the19th Century Knowledge project, where we have convertedhistorical ontologies, specifically the 1750 classification sys-tem underlying Ephraim Chambers Cyclopaedia [6] and the1910 version of the general domain, Library of CongressSubject Headings (1910 LCSH) into linked-data. With regardto scale for the dereferenced resource, the 1910 Library ofCongress Subject Headings contain over 29,000 entries used inthe dictionary catalogs of the Library of Congress published in1910–1928. Figure 1 illustrates a sample catalog entry for theArmories entry. The use of these historical ontologies providesinsight into historical collections in ways not possible withcontemporary systems [7], [8]. A critical goal underlying thework, and the focus of this paper, is to establish resolvable,Persistent Identifiers (PIDs) for historical ontologies used forarchives.This paper presents our initiative exhibiting a preliminaryuse case with Archival Resource Keys (ARKs) as PIDs forthe 1910 LCSH terms. The use of PIDs is crucial for theapplication and sharing of historical ontologies across diversedigital archives and, ultimately, for supporting deeper and
ABLE I1910 LCSH
INDEXING RESULTS FOR THE TH E D . E NCYCLOPEDIA B RITANNICA ENTRY ON SLAVERY AND M ARK T WAIN ’ S “L ETTER TO P AMELA
A. M
OFFETT ON S EPTEMBER
4, 1874.”Digital Archival Resource Terms exclusive to 1910 LCSH vo-cabulary:7th Edition Encyclopedia Britan-nica Article on “Slavery” (19 C.Knowledge Project) Houses; Carries;
Man ; Negroes ;War; Government; Accounting;Accountants; Law; Societies; AgeLetter to Pamela A. Moffett onSeptember 4, 1874 (Mark TwainCollection)
Idiots ; Imbecility ; Turning ;Lawyers; Commons; Schools;
School ; Lays;
Fall ; Judges;Asylums; Building more extensive computational archival science. We supportthis specific effort by integrating ARKs into the Helping Inter-disciplinary Vocabulary Engineering (HIVE) ontology server.The next section provides an overview of preliminary workwith historical ontologies and their relevance with compu-tational archival science, followed by the case for ARKs.We then present the goals of our work and provide detailson our methods. Next, we present our initial implementationexhibiting the approach to serve as the basis for furtherevaluation beyond this preliminary study. The paper wrapsup with a contextual discussion of our work to date, and theconclusion summarizes current progress and identifies nextsteps. II. O
NTOLOGICAL A PPLICATIONS
As more digital archival projects are made accessible forscholarship, the use of historical ontologies for describingthese resources are scarce. One exception is research withthe Helping Interdisciplinary Vocabulary Engineering (HIVE)application and the Nineteenth-Century Knowledge Project(19th C. Knowledge Project) [8].The 19th C. Knowledge Project is building an extensive,open, digital archival collection to support the study of knowl-edge and its transformation. Four historical editions of theEncyclopedia Britannica (3rd, 7th, 9th, and 11th editions),spanning 1797 to 1911, have been digitized for computationalstudy. The combined data set includes over 100,000 encyclope-dia entries, covering a broad range of topics. The project usesboth the contemporary LCSH and the 1910 LCSH for topicalrepresentation. The use of both ontologies provides morecomprehensive access to the material and facilitates study ofknowledge over time. Table I illustrates this difference for anencyclopedia entry and a letter drawn from the Mark TwainCollection [9], where the terms in boldface indicate entriesthat no longer exist in the contemporary LCSH vocabularyas an exact match, indicating concept drift in the controlledvocabulary over time.More recent research has compared the 1910 LCSH withOCLC’s faceted rendering of the contemporary, 2020 LCSH,known as FAST [10]. A sample of 90 full text entries, drawnacross the four historical editions of the Encyclopedia Britan-nica, through the examination of 886 indexing results, foundthat the 31.3% terms generated with the 1910 LCSH did not appear in the contemporary FAST automatic indexing results.Further analysis found that 6.2% of the total 1910 LCSHresults no longer appear in the full 2020 FAST vocabulary,representing temporal concept drift.This initial research confirms that computation-ready his-torical ontologies can provide insight into the archival recordwith access to language absent from contemporary systems.Additionally, a computation-ready historical ontology has ap-plicability well beyond the 19th C. Knowledge Project andenables comparison of collections as demonstrated with theTwain Project letter. The conversion to linked data is an im-portant first step but significantly limited without the supportof reliable persistent identifiers supporting provenance trackingand preservation. The significance of persistent identifiers fordoing this work has been emphasized in the life sciences [11].This need underscores the case for computational archivalscience in the context of ontologies and informs our workadopting the ARK identifiers.
A. Archival Resource Keys (ARKs)
ARKs (Archival Resource Keys) are high-functioning iden-tifiers that provide references and descriptions for objects ofany type [12]. The ARK specification is in active development.Generally, an ARK is a special kind of URI that connectsusers to the named object, its metadata, and the ARK serviceprovider’s promise about its persistence [13]. An ARK is highfunctioning in the sense that it provides access to not only theobject and its metadata, but a specific, revealed commitmentstatement from the institutional provider responsible for theresource relating to access permanence.
1) Persistence:
Impersistence is a major problem for linkeddata, with the average URL remaining valid (not returning a404 Page Not Found error) for only 44 days [14]. In response,a number of PID schemes have been proposed (ARK, DOI,Handle, PURL, URN). While there is no guarantee that thepayload to which any PID points is currently, ever was, and/orwill remain accessible [15], a distinguishing characteristicof the ARK scheme is to provide a standard way to linkfrom an object to a promise of stewardship for it. Combinedwith traditional reputation mechanisms, the details of thatpromise give link recipients the best basis on which to forecastpersistence.
2) Format:
An ARK identifier is carried in a UniformResource Identifier (URI) and shares some of its structure. Assuch it begins with the Name Mapping Authority (or hostname— see Figure 2) that would provide name resolution service(resolver) for the ARK, however, the core immutable identityof the ARK is actually independent of the hostname.The Base Object Name (Figure 2) is composed of thelabel to signify an ARK ( ark/: ), a Name Assigned Au-thority Number (NAAN, e.g., 99152), and an Assigned Name(e.g., 5p30086k). The character “/” after that name indicatescontainment. For example, b41910/5p30086k asserts that theobject 5p30086k is contained by the object b41910. Whileany alphanumeric string can be used for the Assigned Name,the specification recommends using a “betanumeric” [12] ig. 2. Anatomy of a URI containing an ARK character set which removes all vowels and the lowercase l(letter ell) to mitigate ambiguity and help limit transcriptionerrors. More than 667 registered organizations have created anestimated 8.2 billion ARKs [16].A Name Assigning Authority Number (NAAN) is a uniqueidentifier reserved for an organization that wants to createARKs. Prepending an organization’s NAAN to its AssignedNames is required to keep the organization’s ARKs globallyunique. The ARK also supports shared NAANs to permitleveraging certain specific immutable semantics: 12345 forexample ARKs (such as those appearing in documentation);99152 for metadata, controlled vocabulary, and ontologyterms; 99166 for agents (such as people, groups, and insti-tutions); and 99999 for testing and development. We use the99152 NAAN in this work so that receiving software canimmediately infer that the identified object is a “term.”To use a shared NAAN without conflicts requires reservinga “shoulder” (a term borrowed from a part of a physical, metalkey), or sub-namespace under a NAAN [17]. Figure 2 displaysthese and other components of an ARK encoded as a URIto be dereferenced with a resolver. By convention, shouldersrepresent a short, fixed extension to the NAAN. For example ark:/99152/b4 .
3) What’s Different About ARKs:
Most ARKs are cre-ated by organizations that tend to publish them based ontheir own resolvers, but ARKs are intended to be decou-pled from any particular resolver. For example, the ARK ark:/12345/x54xz321 might be resolved with the Name-to-Thing resolver (currently administered by the CaliforniaDigital Library [17]) at the online hostname n2t.net anddereferenced at https://n2t.net/ark:/12345/x54xz321. However,the concept of providing the identifier without a hostname (i.e., ark:/12345/x54xz321 ) is an expected usage pattern thatallows the identifier to persist and be supported by successoror even parallel name mapping authorities.ARKs are atypical compared to other PIDs in that thereare no fees associated with obtaining a NAAN and publishingARKs, the options to specify metadata are flexible by design,and resolution is decentralized, despite a recommended butnot required resolver (N2T) [15]. This is in contrast to otherPIDs like DOIs [18], Handles [19], PURLs [20], and URNs [21], all of which have requirements for centralization, rigiditywith respect to metadata requirements, cost, etc. In addition,all of these schemes ultimately have demonstrated little effecton persistence [13]. A founding principle of the ARK schemeis that persistence is a matter of service and that there must bea way to tie a promise of persistence to a provider’s reputationin that area. Persistence is achieved through a provider’ssuccessful stewardship of objects and their identifiers [20].
B. Helping Interdisciplinary Vocabulary Engineering (HIVE)
HIVE is a linked data, automatic indexing metadata applica-tion (Figure 3). HIVE enables researchers to leverage existingontologies and automatically generate well-formed, standard-ized metadata for digital resources. One of the key features ofHIVE is that it supports the use of multiple ontologies duringa single metadata generation sequence. A curator can workwith both historical and contemporary ontologies in a singleautomatic indexing sequence while also including ontologiesrepresenting different domains. For instance, a curator mayseek to use both the contemporary LCSH and a geospatialontology [22].HIVE’s indexing algorithm identifies candidate keyphrasesin digital sources and then compares them to one or moreuser-selected ontologies in order to suggest a set of termsfor content representation. The HIVE system supports curatorsin ultimately selecting the best ontology terms, although theoutput can also be automatically assigned to digital content,an approach being pursued with the 19th Century KnowledgeProject.Integration of the ARK Identifier schema to historical on-tologies can address such shortcomings and improve discoveryand access to cultural heritage resources. The attribution ofreputation related metadata may help to mitigate the currentstate of affairs where no system of authority files for linkingdata to historical ontologies exists.III. M
ETHODOLOGY
As directed by the ARK FAQ, we registered a NAAN(13183) for the Metadata Research Center at Drexel University(MRC) and a shoulder (b4) to designate vocabularies on theexisting shared NAAN (99152). A primary focus of the work ig. 3. The HIVE Vocabulary Server implements a client-server, web-basedinterface to provide access to a selection of vocabularies and linked data. described in this paper was to resolve identifiers under ourNAAN (13183), supplemented with ARKs specifically forontology terms using the 99152/b4 shoulder.After having reserved our organization’s NAAN, we con-sidered multiple variations of path structures to plan forextensibility beyond the use case we describe in this pa-per. In our initially considered variation, we thought to en-code the vocabulary name into the Assigned Name, e.g., ark:/13183/lcsh1910/5p30086k (where lcsh1910 isthe name of the vocabulary). This configuration, however, doesnot follow ARK best practices (“betanumeric” characters and“first-digit convention” as per the Shoulder FAQ [16].A second structure we considered was to assign one nameto each vocabulary under the NAAN, e.g., b412345. Thiswould produce ARKs like https://n2t.net/ark:/13183/b412345/lcsh1910/5p30086k. This configuration incorrectly implies thatb412345 names an actual object with sub-objects.Ultimately, we settled on the configuration where the por-tion of the ARK indicating the lcsh1910 vocabulary is notsemantically encoded but still represented uniquely. Ratherthan b412345 representing all vocabularies, the b4 shoulderbecomes the root of a namespace and an appended substring(e.g., 1910) indicates the LCSH 1910 vocabulary.The choice to use “b41910” instead of the more human-readable “lcsh” is more opaque (widely seen to benefit per-sistence), more textually efficient than the alternatives weconsidered, leverages a shoulder for future vocabularies, andallows the NAAN to be reused for ARKs that refer to otherthings by changing the shoulder. A downside of the b41910assigned name is that other vocabularies might not have tem-poral semantics (like the LCSH vocabulary using 1910), thusintroducing a schematic discrepancy between vocabularies.Also, the numerical value still encodes some semantics, i.e.,suboptimal opacity.IV. I
MPLEMENTATION
As an initial prototype and proof-of-concept for the ap-proach, we sought to leverage the advantages of the ARK ecosystem, publish the vocabularies online, and connect theidentifiers to the vocabulary payload to make them accessible.One ultimate goal of this process is to make this vocabularyaccessible online to the software implementation of HIVE. Inthis section, we provide details on this process.
A. Establishing PIDs
Our initial approach in establishing ARKs to serve as persis-tent identifiers for vocabularies takes into account scalabilityand extensibility with an emphasis on persistence. The ongoingprocess is progressing from both ends to meet in the middle,i.e., establish PIDs, progress toward resolution, strategicallypublish the payload (vocabularies), and work towards a reliablemeans of access.The Noid (Nice Opaque IDentifier) utility can mint (gen-erate) transcription safe, unique strings that adhere to thepreferred betanumeric recommendations for ARKs [23] sup-ported by the N2T.net resolver and the ARK standard. The tooland/or service is used by a variety of institutions inclusive ofthe Internet Archive (NAAN 13960), University of CaliforniaBerkeley (28722), National Library of France (12148), PorticoDigital Preservation Service (27927), and Smithsonian Insti-tution (65665). A comprehensive, updated list of NAANs isavailable online [24].Resolving our vocabulary entries involves a URI redi-rect chain for which an ARK acts as the commonthread. As an example, upon being provided an ARKlike ark:/99152/b41910/5p30086k , a client or system(henceforth “user”) wanting to leverage the vocabulary couldresolve the ARK. Through MRC being responsible for theshoulder b4 under NAAN 99152, this URI directs the userto https://id.cci.drexel.edu/lcsh1910/5p30086k for further rout-ing.
B. Publishing the Vocabulary
As an additional assurance of persistence that forms oneof the essential elements of a provider’s commitment tostewardship, this URI is not redirected using HTTP but rathermapped using DNS to GitHub Pages, the contents of which re-side in the repository in https://github.com/metadata-research/vocabularies/lcsh1910. The opaque ARK identifier wouldultimately redirect to https://github.com/metadata-research/vocabularies/lcsh1910/5p30086k for the period of time thatGitHub hosting is used, and would be updated in the futurewhen the hosting arrangement changed. MRC is responsiblefor the response to any requests as forwarded from a resolverfor the shoulder b4 under the NAAN 99152. This designchoice helps to ensure data accessibility beyond the confines ofa single provider while avoiding possible user misconstructionof the ARK identifier. Further, hosting the vocabularies onGitHub encourages version tracking of any vocabularies thatare still progressively being generated (e.g., on contemporarytopics) or refined (for historical ontologies from perhaps notborn digital sources) to facilitate accountability, openness, andcommunity engagement. . Incorporating ARKs into HIVE
The current implementation of HIVE defines a unique URLfor each term in the 1910 LCSH. Our initial prototypesand mock-ups use the shared NAAN for display purposes.These URLs will be replaced with unique and opaque ARKidentifiers. The Noid utility is used to generate (mint) a uniquecore identifier that includes the NAAN and shoulder value.This identifier is then embedded in a URL to form the ARKidentifier, for example, ark:/99152/b41910/5p30086k .The 1910 LCSH database in HIVE will then be regeneratedto incorporate ARK identifiers for each vocabulary term.Figure 4 illustrates the HIVE web page for the Abbeysentry in the 1910 LCSH. The value for the URI/URL hasbeen replaced with a URI/PID field containing a uniqueARK identifier. The “Related” entries (Cathedrals, Convents,Monasteries) will also be assigned ARK identifiers and linkedto their respective terms.
Fig. 4. A catalog entry for the LCSH1910 vocabulary.
V. D
ISCUSSION
The case we present in this paper and our approach ofuniquely identifying a historical vocabulary using the ARKstandard and subsequently, an ontology exploration tool, wasa first, exploratory step in determining the usefulness, feasi-bility, and merit of making the data available using emergingstandards. The ARK ecosystem is evolving (i.e., currently anIETF standards draft [25]) and thus exploration of this sortcan serve as a basis for practice by others that would like toaccomplish similar goals.While using the Git distributed versioning protocol (viaGitHub) as a repository for data storage does not ensure per-manence, we are considering additional strategies for redun-dancy. However, the integration into a system that inherentlyinstills conventional distributed version control may provideadditional use for vocabularies that are still evolving, unlikethe static LCSH1910. Contrarily, the public availability of thedata might further facilitate use and reuse of the data itselfbeyond implementations we can initially anticipate in thispaper.There is an apparent nascent interest in historical ontologies,such as the use of multiple period vocabularies to contextualizenineteenth-century sideshow performer images in the RonaldG. Becker Collection of Charles Esenmann Photographs [26]. Another example is the PeriodO Project, a collaboration be-tween University of Austin, Texas and University of NorthCarolina, which has explored the availability of historicalterms for representing time-periods, using its own “shoulder”, ark:/99152/p0 , to represent historical concepts as linkeddata [27]. Advancing towards a more extensive environment inwhich entire historical ontologies can be located and appliedto archival collections, concepts need to be resolved usingpersistent identifiers such as ARKs.
Fig. 5. Registering a NAAN requires the organization to commit to datapersistence by checking a box.
Persistence as a chief motivator for this work is coupledwith commitment to stewardship. At this time, persistence inany PID scheme cannot be enforced, and is usually approachedby a combination of up-front declaration of intention (e.g., seeFigure 5), broken link monitoring, reporting, and sometimescommunity intervention (e.g., the Internet Archive steps into rescue content from an collection that has lost funding).Our initial plan to use Git is a step toward this and GitHub’scommitment to ensuring long term preservation is demon-strated via the recent Arctic Code Vault endeavor. However,limitations surface when considering how often updates willbe made as well as threats posed by climate change, wherethe future existence of the arctic is at risk.While ARKs are protocol agnostic, the current expectationis for the resolver to be accessed via HTTP to dereference theARK embedded within a URI. However, the scheme/protocol(qua RFC3986 [28]) is not embedded in the core immutableARK, allowing the dereferencing procedure to potentially useanother system beyond HTTP for ARK resolution. This aspectof portability allows ARKs to be relevant to systems beyondthe Web. The initial work we investigated and performed inthis paper has helped our team to more fully appreciate andunderstand the immense flexibility and durability of the ARKecosystem and anticipate future use cases. This may provesignificant for ontological systems to operate in other elec-tronic and analog environments – critical for the persistenceof historical vocabularies.In manifesting the key archival principles of evidentialand enduring value [29] and grounding the work presentedhere in the context of computational archival science (CAS),the value of ARKs for tracking provenance and promotingpersistence are key. While this case study focuses on the1910 LCSH, the work with ARKs has implications for afull range of historical ontologies, many of which remainin analog format. The implications of this research extendbeyond LCSH and humanities resources, and demonstratehe potential for converting other historical ontologies intolinked data, facilitated by ARK identifiers. This approachcould potentially be applied to scientific, medical, agricultural,and other historical terminologies.VI. C
ONCLUSIONS
In this paper, we present a case exploring the appplicationof ARK persistent identifiers for the 1910 Library of CongressSubject Headings by registering our own shoulder under theshared NAAN for terms and investigating an approach forestablishing ARKs in a systematic way. Our initial prototypesand mock-ups in the context of the HIVE tool allow forfurther integration of other vocabularies and provide a scal-able approach for integrating PIDs into HIVE in the future.Facilitating the availability of linked historical data throughthe efforts described in this paper is a first step in preparinghistorical ontologies for computational archival science in awide variety of contexts.VII. F
UTURE W ORK
In our initial exploration of using ARKs for a historicalontology, we sought to facilitate persistence of this data bypreventing reliance on a single service or data source. For thetime being, the N2T resolution service is primarily responsiblefor resolving ARKs. In the future, we hope to establish amethod and practice of propagating the resolution of ARKsto an alternative data source to encourage more distributedARK resolution.This initial effort was an exploration of the applicability andfeasibility of utilizing an existing PID system as a means ofresolving and facilitating access to the LCSH vocabulary. Weexhibited the feasibility of systematically integrating ARKswith a historical vocabulary and anticipate refining our pro-cess. We also plan to apply our initial process to other relevantvocabularies.This initial exploration proved valuable as a low-risk,preliminary exercise in the potential to integrate PIDs withhistorical vocabularies. We learned a lot about both systemsthrough this exploration and realize that there are still manyunknowns whose challenges will only be surfaced throughexploration and prototyping of this nature. The long-termimpact of making historical ontologies available with PIDshas implications not only for historical documents in thehumanities but in enabling the computability of archives. Thereis more we can learn from archival practice in the areas ofprovenance, life-cycle management, use, and reuse that canbe applied to big data. R
EFERENCES[1] R. Marciano et al. , “Reframing Digital Curation Practices througha Computational Thinking Framework,” in , Dec. 2019, pp. 3126–3135.[2] N. Payne, “Stirring The Cauldron: Redefining Computational ArchivalScience (CAS) For The Big Data Domain,” , 2018.[3] J. Chen, W. Lu, and O. Zavalina, “Workshop on Organizing Data,Information, and Knowledge in Big Data Environments,” in , Jun. 2019,pp. 459–460. [4] M. L. Zeng, “Smart Data for Digital Humanities,”
Journal of Data andInformation Science , vol. 2, no. 1, pp. 1–12, Feb. 2017.[5] W. L. Hamilton, J. Leskovec, and D. Jurafsky, “DiachronicWord Embeddings Reveal Statistical Laws of SemanticChange,” arXiv:1605.09096 [cs] , Oct. 2018. [Online]. Available:http://arxiv.org/abs/1605.09096[6] E. Chambers,
Chambers’ Cyclopaedia , 1750. [Online]. Available:https://artfl-project.uchicago.edu/content/chambers-cyclopaedia[7] S. Grabus, J. Greenberg, P. Logan, and J. Boone, “Representing About-ness: Automatically Indexing 19th- Century Encyclopedia BritannicaEntries,”
NASKO
FAST: Faceted Application of SubjectTerminology . Libraries Unlimited, 2010. [Online]. Available: https://products.abc-clio.com/abc-cliocorporate/product.aspx?pc=F2275P[11] J. A. McMurry et al. , “Identifiers for the 21st century: How to design,provision, and reuse persistent identifiers to maximize utility and impactof life science data,”
PLOS Biology
Journalof Library Administration
Proceedings of the Association for Information Science and Technology ,vol. 52, no. 1, pp. 1–8, 2015. [Online]. Available: https://asistdl.onlinelibrary.wiley.com/[23] “NOID.” [Online]. Available: https://n2t.net/e/noid.html[24] University of California, California Digital Library (CDL), “NameAssigning Authority Number (NAAN) Registry.” [Online]. Available:https://n2t.net/e/pub/naan registry.txt[25] J. A. Kunze, “The ARK Identifier Scheme Draft 24.” [Online].Available: https://datatracker.ietf.org/doc/draft-kunze-ark/[26] B. Dobreski, J. Qin, and M. Resnick, “Side by Side: The Use of MultipleSubject Languages in Capturing Shifting Contexts around HistoricalCollections,”
NASKO