Notes on Meetings at ALA Annual, New Orleans, 2018 with a Focus on MARC and Other Encoding Standards, and Linked Data (Plus a Couple Other Meetings of More General Interest)
Submitted July 12, 2018, by Jim Soe Nyun, Chair, Encoding Standards Subcommittee
MARC Advisory Committee I
Session 1: 6/23/18
Sessions 2-3: 6/24/18
Meeting agenda with links to papers: https://www.loc.gov/marc/mac/2018/2018-02.html
Announcements of latest fast-track changes:
- Added $4 to 730
- $r redefined in Field 382 (MLA request)
Proposal No. 2018-02: Subfield Coding in Field 041 for Accessibility in the MARC 21 Bibliographic Format
- Amended to clarify wording of new subfields $p and $r, passed unanimously
Proposal No. 2018-03: Defining New Fields to Record Accessibility Content in the MARC 21 BibliographicFormat
- Amended to drop $0 and $1 as there was not a direct use case that people could see; added $8 to both fields; revised wording of the field’s scope and for indicator 2
- Wrinkle for music: No clarifying statement that musical notation is to be understood as text was written into the final text for 341/$a: “Mode required to access the content of the resource without using assistive features (one of textual, visual, auditory, or tactile).” Best practices could help us.
Discussion Paper No. 2018-DP10: Designating Access to Online Resources in Field 856 in the MARC 21Formats
- Roy Tennant presented for OCLC. Some general background information:
- 50 million records in OCLC could likely be processed by algorithms to supply an indication that a resource in an 856 is open access
- Only .12% of bib records have multiple 856 fields
- Setting a value for “hybrid open access” could require record maintenance
- There was interest in seeing this paper return as a proposal, but the room felt that the discussion paper was not well enough developed to convert to a proposal on the spot
- There was also general agreement that the 856 field is full of ways to record ancient technical information that hasn’t had any relevance for many years. There was a hope that someone would undertake this, but it won’t likely come out of a proposal based on this discussion paper
Discussion Paper No. 2018-DP11: Open Access and License Information in the MARC 21 Bibliographic and Holdings Formats
- This paper has many relationships to the preceding one (DP10), and once again the current not-useful, not-current state of the 856 field came up. That task would be too large for any revision of this discussion paper, however.
- In the end it appeared that developing Field 540 had the most support, and the paper may return as a proposal that develops this option
Proposal No. 2018-04: Versions of Resources in the MARC 21 Bibliographic Format
- Amended to add $3, $6 and $8 to Option 2, define new Field 251. Passed unanimously, with the acknowledgement that wording is not perfect and that the field may need to be developed as terminology and supporting vocabularies evolve or are established.
Proposal No. 2018-05: Multiscript Records Using Codes from ISO 15924 in the Five MARC 21 Formats
- The committee preferred the option to use $6 and to develop the language in Appendix A of all five MARC formats.
- One abstention, with the rest of the committee voting in favor of passing the proposal.
Discussion Paper No. 2018-DP07: Designating Sources for Names in the MARC 21 Bibliographic Format Presented by Chew Chiat Naun
- There was general consensus that the basic idea was sound, but many different opinions on what a proposal based on this paper should look like. Some of the issues:
- The paper looks at fields where names are recorded, but these same fields are often used to record “titles” or “works”
- The PCC Task Group on URIs in MARC was mainly interested in dealing with names, but it’s hard to separate just names from what we could encounter in these fields
- There are potential problems with a name coming from one authority and the title from another (or not at all, e.g. 100/240)
- Should $2 be repeatable?
- “$2 local” can be problematic
- “$2 viaf” is not authoritative in citing an authority since VIAF aggregates multiple authorities, and might make more sense as a real world object
- The paper may return as a proposal or further developed paper, but there will be a lot to figure out.
Discussion Paper No. 2018-DP08: Use of Field 024 to Capture URIs in the MARC 21 Authority Format
- Option 2 was favored (e.g., developing $0 and $1, but leaving $a for non-URI and non-RWO identifier strings)
- Will likely return as a proposal
Discussion Paper No. 2018-DP09: Improving Subfield Structure of Field 245 in the MARC 21 Bibliographic Format
- Jay Weitz presented this paper, which was developed mainly by OCLC’s Leiden office
- General agreement that the 245 is not clear in expressing all the elements that are placed there,with the inability to repeat subfields causing many problems. However there was little agreement on how to improve it. Discussion revealed many issues, including:
- Neither of the two options presented had general acceptance
- Examples oversimplified the reality of how complex titles can be
- Confusion over how to treat parallel title statements versus titles that consist of multiple title, other title, and responsibility statements
- We could not rely on machines to reliably modify existing 245 statements retrospectively
- Ultimately much reticence to take on this giant task. A straw poll in the room showed 8 people who were interested in developing this paper further, versus 12 who thought this was too big or complex to take on at this late stage of MARC.
OCLC Linked Data Roundtable 6/23/18
[…meeting more than ½ over after a long bus ride and walk to opposite end of convention center…]
Arrive at the end of the second presentation:Jeremy Myntti: Presenting on the Western Name Authority File https://www.imls.gov/sites/default/files/lg-72-16-0002_proposal_narrative.pdf
DPLA API being queried
Exploration of where to go next
Non-Latin script resources in BIBFRAME (Sally McCallum)
Currently LC practice in bib records has been:
- Latin only in non-880 fields
- Non-Latin entirely in the 880
- Transliterated data for Main entry, titles, imprints, notes, etc.
Authority has Latin in 100 but non-Latin may be used anywhere else. Problems: Much duplication of effort, transliteration issues including what transliteration scheme was used.
LC’s BF WORK practice: transliterated title, names are transliterated or linked; optionally may include non-Latin variant titles in work record or non-Latin script.
All instances link to work, no redundancy for INSTANCE, local data kept with instance description. Some savings in not having to Romanize. Full Unicode available, more scripts.
There will be a presentation later today on this topic. Reference ALCTS Non-English Access Working Group on Romanization Report 12/15/2009
Jessica [?, first presenter] on using wiki data input for non-Roman scripts. Actually pretty easy using her interface. The tool is good; problems are conceptual.
Problems with Olmeca-S: Can publish in JSON-LD but only a record at a time. Some issues with searching simply. Reconciliation issues, ways to work with OpenRefine? Currently separate datasets are beingexported and worked on.
Any mapping to mapping to MARC, particularly en masse? No work yet. Want to look to implement in the future.
To Jeremy: What will be the model for regional authority hubs, going local to national? Now they send local name candidates to someone to add NACO records. Have done about 20 names but have about500 more they could think to contribute. (Possibly an issue here of scaling up.)
Work records capture variants, question on whether instance variants could feed up to the work record? Not expecting to automatically populate work records from transcribed instance information.
Open question to Sally McCallum about outgrowths from BF work: They’re developing efficiencies with building out modeling, displaying information from linked records.
Question to Jeremy: How would they integrate multiple AFs? Query both LCNAF and Western Names File. Federated search. Sally thought adding URIs to MARC could lead to enriched BF content. People mentioned models with less duplication, parallel statements with and without parallel information.
[Western names question:] After names are reconciled will EADs across repositories be synced up with the unified forms? Not clear whether partners are willing to do this synchronization. Grant runs through end of October. Maybe a subsequent grant.
Question about linked data pilot project, creating descriptions for translators, where they would not transcribe authors, only translators? Author name might be in a fingerprint, but not in the description,proper. Some questions about whether a translation would be a new work under WEMI.
MARC Formats Transition Interest Group 6/23/18
Chew Chiat Naun
PCC Task on URIs in MARC
- Background/history of group
- Basic concept was that some of our structured data is available in external resources, e.g. VIAF. It would help to include these links if we made them available in our records
- Faceted vocabularies have become more available and important
- Steven Folsom’s 2015 Best Practices for Linked Data in MARC, was an instigator
- Worked with various players including Casalini, RSC, implementers of URIs in MARC, others
- National (BL, German libraries) libraries have had an interest in this
- Did not intend to make MARC a carrier for linked data. MARC can only go so far, and the intention was not to give MARC eternal life. Still, “MARC is not going away any time soon.”
- Some work looked at existing MARC to see how well it could move easily towards linked data. 7XXs are fairly compatible, but others not. Talked about RWOs versus authority control. We can’t assume the contents of MARC fields have a unitary vocabulary or authority, so $2 in these fields could be useful. Figured out later in to the work that the headings don’t point to entities.Added the use of $4 for a relationship
- Work identifiers, those algorithmically generated by OCLC’s project, other sources
- Papers will be heading to MAC to address some of the above
- Look at outcomes page for useful documentation:
- Real World Object, $1, $0, position paper on expanding the use of URIs in authority records. Best practices group being formed
- Terry Reese pointed out that he can make some of the MarcEdit work available to otherdevelopers
- Next: reach out to practitioners to see what they’re doing and what they need. Look at use oflinked data vocabularies. How can WorldCat identifiers be used and preserved? Where will this work live once done?
A long and Winding Road to Linked Data
Experimenting with Linked Data at NLM
- NLM’s work to add machine dereferencable URIs in their records. Showed some vignettes in their efforts to implement the above in relation to the PCC task force
- They joined the BF experimenters group, then an early implementer group. NLM formed amulti-dimension linked data initiative. Published MESH in RDF. Experimenting with SEO in their digital repository. Involved in PCC linked data working groups
- How has the TF’s work intersected with NLM’s? Collaborated on some MAC papers. Terry Reese had started work on MARC Next tool, including tool to add URIs to MARC. It proved you could do this programmatically. Resulted in MAC paper to add $0. Terry added MESH RDF to the lookup
- Looked at fields to which $0 applied in data fields, which subfields were part of the label
- NLM wanted to move from theory to actually doing something in RDF. They added identifiers for MESH into MARC records. Was a harder task than originally though. 2017 NLM added MESHURIs to their distribution records
- Digital repo work: Didn’t want to work with strings. MARC Next tool opened up working with NLM records. All their repo items start out life in MARC. Schema.org>>JSON LD serialization added to their repo resources
- Casalini had interesting projects going on. Share VDE hs good components, can pull data from various places such as Wikidata, has a working interface, etc. Casalini became a project participant, was interesting that they had MESH. Looking for ways to make MESH URIs more easily available
Considerations related to changes:
- Making changes through the MARC Advisory Committee is slow, and it can take a while to makea change to MARC
- Institutional priorities don’t always align with moving ahead
- Technical limitations, e.g. not everyone has a triple store
Lack of funding is not entirely an impediment. NLM has developed MESH in RDF. Gets some educationfor funding. Their government mandate says they can’t put money towards developing tools. None oftheir work above was done with external funding. They look for possible collaborations, e.g. PCC URI TF,looking at projects out there that they could work with. They limit what they take on. Other resources:cheap coursework, lots online, work with interesting people.
- Still working on going from MARC to Schema/JSON-LD. SEO is capricious. SEO-enhanced toolshave received more links, and it could expand exposure
- Answer to question about doing things without funding: SHARE VDE had no cost for Phase I. They still have a voice in development even though not part of subsequent formal stages, however
- Are vendors being part of the discussion? Not really at this point. The work was to see whatvendors can do. We might need to be clearer with where things are headed before committingto developing products. OCLC has been involved, partly to build on what they can provide. Somevendors were part of information sharing early on. Most vendors now offer URIs in records forfree now
- At least one vendor (for Primo product) apparently uses identifiers in records to minereferences. Apparently is causing some conflicts, [though I was not totally clear of the problembeing stated]
Big Data in Libraries: Friend or Foe? 6/23/18
Intellectual Freedom RT
William Marden, NYPL
Peter Brantley, UC Davis
Erin Bergman, San Jose Public Library
100 seat hall, all full…
How do you define BIG DATA?
- Peter, terabytes to petabytes of data, big amounts of data from systems or sensor arrays. Not allproblematic data is only BIG. Erin, echoing Peter’s comment, lots of information run throughvarious algorithms.
Is there a diff between big data and surveillance?
- Sometimes it can be the same. When collected anonymously maybe less of a problem. What’scollected now may in the future turn into more focused, less anonymous data collection.“Reidentification” is something to be guarded against, mainly coordinating one dataset againstanother. “Trusted data environments” could be a safer way to go. Safest practice is to collectonly information that is being collected; look at how data could be used or misused. Considerthe dataset immediately compromised. GDPR is a good direction towards what can be collected.
How do you decide how much to track?
- Best not to track individuals. Be careful about trusting that something is being madeanonymous. Be careful about what data from services you take advantage of, e.g. credit scores.Maybe you can ask a user to opt in to doing some analytics on their online behavior. Lessonsfrom the recent European implementation of GDPR: toss what you don’t absolutely need.Campuses might do analysis looking at library users. Don’t use data collection as a substitute forinteractions; connect with the users. Requiring opting in to agree to user terms is NOT an optionfor public libraries. It’s not true sampling if we’re working with opt-in strategies, populationversus sample analysis. Some data will still end up being collected.
- Libraries are among the most trusted institutions right now. People want control of theirinformation. Every state has a statute protecting library information.
Is it time for different legislation that protects library data?
- Control data ourselves as much as possible; you don’t know what a 3rd party vendor would dowith data. UC doesn’t have a system policy on patron data, probably could use one. Don’t trackthe reading habits of patrons.
Who should be the guardian of patron privacy?
- E.g. no business use for computer logs. Libraries sometime shave non-library staff who aren’tup on all our privacy issues. Some organizations have a chief privacy officer, requires deeperpockets. Library needs to be involved in privacy issues for larger entity (e.g. city).
- Beware of ILSs. Beware of profiling or recommending systems. Point to ALA privacy guidelines, anational standard for library vendors. Authenticating via services like FB or Twitter or includinglike buttons can compromise patron data. Many cool toys, but we can’t always be cool. Who arethe data beneficiaries when a company folds?—be sure to have that condition negotiated in acontract.
Faceted Subject Access Interest Group 6/23/18
FAST update (Jody DeRidder, OCLC)
Survey from various institutions went out last year. Most responses from US, mainly from tech servicesand cataloging Interviewed 10 institutions to follow up on survey (BL, Brown, Columbia, Cornell,Harvard, Penn State, Stanford, Yale, Analysis & Posicy Observatory and Royal Melbourne Institute of Technology, the latter two in Australia).
- Needs FAST supported long-term.
- OCLC is loading FAST terms into the the authorities database.
- Needs to improve current tools, enable batch conversions, and APIs.
- Provide ongoing FAST support.
- Don’t make FAST too complex.
- Some confusion about when to use fields.
- How to request new and alternative terms. What is the relationship to LCSH? How to deal withterms that are from different domains.
- Time periods are US-weighted, and may not be clear to other users in other countries.
- Different terms have different meanings internationally. (E.g. “Health & Hygeine” in Australia).
- If we have new terms how should the relationship to LCSH be viewed. Sole source? Workthrough SACO? Break ties? Treat LCSH as one vocabulary.
- Not ready for a break. But LCSH is often inadequate. What are out options?
- SACO? Community input mechanism? Use external vocabularies?
- Incorporating terms could make FAST easy to use.
- Requests for parallel culturally-appropriate terms, multiple languages. WikiData might havesome options, particularly multi-language capabilities.
- Complexities of adding multiple sources, mapping among terms.
- Need more community engagement, a transparent process, a fast process.
- Community voting? Working groups to address various parts of implementation anddevelopment. Establish editorial policies.
Next steps to develop editorial policy committee: scope of work, structure, requirements forparticipation, selection, terms…
Batch Authority Searching with Python (Kelsey George, UNLV) [Recorded presentation] “Very much awork in progress”
Worked with ETDs from ProQuest; destination was their IR.
PQ sent Excel spreadsheet with metadata.
The above works with PQ MARC records that have their own scheme of assigning terms.
Went from records using MarcEdit to OpenRefine. Cut and pasted the facets into a single column thende-duped. Ended up with 257 unique values that she fed into OCLC’s search for authorities. Came backwith 11 successful searches, 20 errors and 226 too many matches. The tool searches all of the OCLCrecord. Ended up doing a manual browse of the terms, and correcting records based on the results.
Took above to FAST converter.
Would it be useful to develop a Python-based toll for long-term work?
Send Kelsey.firstname.lastname@example.org if you have answers or ideas.
Open discussion on the use of vocabularies designed for faceted subject access (moderated by LynnGates and Lucas Wing Kay Mak) [No comments to moderate. Meeting adjourned]
Metadata Interest Group 6/24/18
The Cataloging Lab: Encouraging Collaboration for a More Effective and Inclusive LCSH Proposal Process (Violet Fox, Dewey Editor at OCLC, as of Monday; this is a personal project, not something from OCLC)
Slides will be up at ALA Connect
Grew out of a Google Doc process to propose a new LCSH term.
Wiki platform grew out of the above.
Talked about the current SACO program and process. It takes 6-10 weeks for one round; may extend if aproposal needs to be revised.
Cataloginglab.org is a WordPress site that enables submitting LCSH headings. A template heading isprepared and then put up on the wiki for comment.
The process is an opportunity for us to be responsible with the vocabulary we use. A group process canhelp reduce individual biases.
- Revisions? (“Transsexuals” came up as a problematic term.) Showed example of headings inprocess, including one where a term “ableism” was added as a reference.
- Demo’ed site briefly. She has de-emphasized MARC to break down barriers to contribution fromcommunities outside cataloging.
- Who submits? Violet Fox right now. But everyone is currently empowered to submit to LCSH.
- Comment from Paul Frank: He’s very encouraged to see projects like this.
MLA Report (see below, at end of the report on this meeting)
CC:DA Report: Currently light in their activities because of the RDA freeze. RDA beta site is up; materials will be put up on the RSC site. Midwinter will have another iteration of Toolkit sessions. Toolkit webinarsare coming up. RDA in its original form is frozen; no further updates.
Tomorrow’s CC:DA will look at impacts on MARC21 development.
LITA/ALCTS Metadata Standards Committee: Recently adjusted its charge from the original moreamorphous charge.
This year the business meeting will look at the transition from two current chairs to new ones.
This afternoon there will be a program on inclusion and diversity issues as they relate to metadata,followed by a discussion where attendees will be welcome to share their ideas.
Voting for officers:
On ballot, Darnell Melvin, vice-chair and Chair Elect Ming yawn Like, programming co-chair; KelseyGeorge and Rachel Tilley Blog Coordinator; Rachel is the winner.
Scott Carlson from DLFDLF AIG Metadata: Google “Metadata Clearinghouse” an attempt to collect application profiles at apoint in time. An attempt to help share profiles for a lot of profiles to look at how others have dealt withthe same materials / the site has things like code and best practices
Brief Report on Metadata Activities of the Music Library Association Report prepared by Jim Soe Nyun forthe June 24, 2018 meeting of the Metadata Interest Group
Background: The Music Library Association has one overarching Cataloging and Metadata Committee, plus three subcommittees, one of which is the Encoding Standards Committee, for which I’m Chair untilFebruary of next year. The other subcommittees have some folio assignments related to link data,particularly vocabulary development and maintenance. The Encoding Standards Subcommittee has itscore assignment MARC and any other schemes that might be used to encode metadata, and as chair ofthe subcommittee I am designate by MLA as a liaison to this group, as well as to the MARC AdvisoryCommittee.
MLA’s Cataloging and Metadata Committee also may establish working groups devoted to a specialproject. Falling into that group has been the Linked Data Working Group, LDWG, which I co-chaired along with Kirk-Evan Billet until last fall to help with the Music Library Association’s participation in thePerformed Music Ontology (PMO) component of the larger LD4P project. As the Music Ontology projectbegan to issue documentation on GitHub and Biblioportal, LDWG was resuscitated with Hermine Vermeijas Chair to review the project documentation and provide comments from MLA as the constituency ofprimary interest in the work of the PMO group. Several papers were reviewed by the group this pastwinter and spring, and comments submitted back to Nancy Lorimer of the PMO project.
In more old-school metadata work, MLA proposed several changes to the MARC format for bibliographicdata that we hope will help to better associate some parts of the record with other data elements of thesame record. The changes to validate subfield $3 for several 3XX data element fields were passed by theMARC Advisory Committee at the time of my last report to this group, and the changes have beenpublished in the recent MARC Update 26, and will be available for use in a few days after the 60 dayreview period concludes. Finally, the Encoding Standards Subcommittee and other members of MLAprovide feedback on papers and proposals being considered at the MARC Advisory Committee meetings,and a consolidated response to the nine current MAC proposals or discussion papers was issued to theMARC listserv in preparation for MAC’s meetings this weekend.
LC BIBFRAME Update 6/24/18
Introduction: Sally McCallum
Introduced speakers and presentations
LC BIBFRAME 2.0 Pilot progress report:
- Working on a more formal report with responses from people working on BR 2.0, will bereviewed and put out soon. Worked with LD4P participants to include their work in the BF test. The BF catalog includes records from the LC catalog that have been converted for testers toconsult. 4 billion triples from ca. 20 million MARC records in the MARC database. Re the BFEditor, it can deal with in-process records, can save. Now can deal with many formats doneunder BF 2.0. Their work includes rare materials which don’t rely exclusively on RDA. Participants at LC worked with their triple store, which was continuously update as catalogingfrom outside the Pilot was converted for inclusion.
- NEWS: The Library is pursuing BF as a viable alternative to MARC.
- Creating and Updating a BF Database Jodi Williamson Loc.gov/BIBFRAME
- 17+ million converted records, 1.2 million uniform title authority records, records added everyday from ongoing MARC cataloging.
- Showed editor tool, includes a left-anchored search
- Match and Merge process looks at 130/240 titles indexed as nametitle, also 1XX+245 for titlematching. Merges work infor from “bib records” and creates instance based on other data in therecords. Some types of works, like music, have special needs, where works should be related tothe larger work, but not necessarily expressions of it. Music got a shout out for all our goodname/title records created, making merging easier. Not all work records link to LC resourceshave instance records since they don’t have everything. More records will be matched andmerged. Name/title works pretty well until you encounter things works titled “Untitled.”
- Editor currently includes work with various catalogers at LC, sound recordings and rarematerials, e.g.
- Showed instance for a map
Things to improve:
- 7XX related works aren’t dealt with well. Now they’re stubs but should link to related works
- Contributors from 7XX tags are sometimes lost in conversion
- Load sequence and system control numbers impact conversion; revisions assign new numbersand links can get broken
- They still need to develop more profiles
- Need to develop a base BSR description that can be expanded
- Need to be able to easily clone a description
- Need to accept different serialization schemes
- Working to see how Casailini’s data could be loaded
- Ingesting CIP and ONIX data
- LC’s BF file is open for exploration
- Keep developing editor
From MARC to BIBFRAME in the SHARE-VDE project (Tiziana Possemato, Casalini Libri) Background onSHARE-VDE, project to move LAM data into linkable data, including MARC to RDF using BF model.Includes editor/interface. A collaboration with the library community.
- Enriching MARC with URIs
- Data publication in BF
- Batch data updating and dissemination
- Development of use cases from the user community
- MARC conversion relies on a Similarity Score, and then “Authify” to identify each entityand enrich with external data. Creates data clusters. Starts with authority records feedsto POSTGRES dB which is processed goes to CLUSTERS KNOWLEDGE BASE.
- Bib records go through similar process for enrichment.
- KB goes through “Lodify” processing then into triplesore with different output.
- Process tries to extract different entities in each record. Reconciliation works withautomated or manual processes. The former allows for large datasets. The result is a“cluster.” Showed Vivaldi examples as a result of going through the process.
- The Lodify process parses MARC, so e.g. a 300 field can go into BF properties (e.g.“dimensions”)
- Phase 2 deliverables:
- Datasets in BF 2.0 format, enriched with “tuples” derived from MARC Clusters in RDF Datasets in BF 2.0 URIs table for external sources
- Followed LC’s work to see what they need. Includes LC BF extensions.
- Lodify is the framework that automates the conversion and publication of bib andauthority data in RDF according to BF 2. LODIFY maturity level test shows good results.
- Part of the goal: “To create a new ecosystem.”
Using BIBFRAME in multi-institutional projects (Jeremy Nelson, Colorado College)
- A Colorado Wyoming DPLA service hub, includes working with various cultural heritageorganizations including Denver Public, Colorado College, University of Wyoming, HistoryColorado, American Heritage Museum, others.
- Input formats in CSV, MODS XML, JSON (not according to a standard). RML (RDF MappingLanguage) was used to go from these sources into RDF. They simplified the BF 2.0 model for themapping work. RML is RDF, RDF to create RDF. They had to make custom mapping’s for customdata models, but MODS XML was more standards compliant and reusable. Uses XPATH to mapXML to BF RDF.
- Output front the triple store can be exported as JSON in the DPLA data model.
- Above is from the BIBCAT open source project. Used, at first, Blazegraph RDF Triplestore butperformance was an issue. Later used RDFFramework, RDF to Elasticsearch with RML mappings.
- Planestopeak.org service hub, using ResourceSync, a replacement for OAI PMH.
OCLC research with BIBFRAME (Nathan Putnam, OCLC)
- What OCLC has been doing with the BF 2.0 converter.
- They reviewed the style sheet to see what would need to be modified for the OCLC businessmodel. Converted 11 million to MARC XML and ran them through the BF converter.
- Work IDs, cluster IDs helped
- URIs are very important
- Edited to deal with gaps re instance records/info
- They Populated 758 with Work/cluster IDs Looked for $0 and $1 Preferred URIs for VIAF andFAST instead of LCSH.
- What will happen with 880s? 880s are paired with the Romanized fields. Instances are innon-Roman.
- LC will do both mapping and the converter tool. LC needs it and there’s public interest. LC hopesthey’ll be out very soon.
- Please share slides, video would be nice.
- How will RDA entities be expressed in BF? Issues with difficulties in data modeling. In BF, e.g., bf:hasExpression can help clarify RDA work relationships.
- Item information for BF is currently being converted. It’s an area where item extensionvocabularies for vendors might be necessary.
- Are the enchanted Casailini records being exported and distributed? The deliverables were to goback to the partner library. One project shared broadly, one with just the contributing library.
- How did LC choose which records were part of their work sample? Records with LCCNs only.
Special session on RDA and MARC
(Notes available at: http://www.gordondunsire.com/presentations.html)
The meeting began with three presentations to expose some of the issues related to moving the new
RDA model into MARCRDA and Data Encoding: The impact of the 3R Project (Gordon Dunsire)
- RDA doesn’t talk about MARC though it does serialization in RDF
- Overview of 3R project genesis, and of current state of Toolkit in Beta (including development ofentities etc. almost complete).
- New entities for agent, collective agent, nomen, place, timespan.
- Some other changes, e.g.: Attributes become relationships, e.g. date of birth > Related timespanof person; more inverse relationships, e.g. Date of birth of [related person of timespan]
- 13 entities, 1700+ elements. Balance changes include more elements for work than formanifestation previously.
- ISBD review will commence soon.
- RDA entities:
- No change for Work, Expression, Manifestation, Place Item always exemplifies amanifestation.
- RDA Corporate Body and Family are types of Collective Agent.
- Person restricted to human beings, does not cover personae, non-human personages.
- Make explicit kinds of recorded data
- CLOSED WORLD ARCHITECTURES:
- Flat file
- Big/authority modeling split
- Relational DB model
- GLOBAL, OPEN-WORLD ARCHITECTURE:
- Recording methods (formerly, 4-fold path) Unstructured description can only be mined forkeywords Structured description (includes e.g., structured notes, controlled vocabularies)Identifier, local numbers, ISBDs, a code for a concept taken from a controlled vocabulary, asexamples. You need to know the domain, the context to make sense of these.
- IRI/URI: globally unique identifiers. LRM says IRIs can be used as an identifier, RDA not.
- Data provenance is important. “Metametadata.”
- RDA in RDF, RDF Resource Description Framework
MARC 21 in RDA Toolkit
- MARC mapping RDA Registry RDF to output is currently very awkward; they’ve developed a CSVfile that you can download from the site
- Inputs: tag, bib/authority, used to map RDA entities to MARC mapping
- Currently no RDF mapping for MARC 21.
- Approximately: RDF element>hasMARCmapping>MARC 21 element
- Currently there is a search that lets you go from RDA element to MARC.
RDA in MARC 21: Accommodating 3R
- Context for change: 3R Project timeline, new RDA entities, new guidance chapters Context forchange in MARC 21: existing, new MARC21 content designation; non-MARC data formats
- Basic principles:
- Choice of MARC 21 format, separate bib authority, other Granularity Consistency?
- New RDA entities
- RDA Entity: No place in MARC to capture current RDA definition. Probably no need tocapture in MARC since this conceptan abstraction
- Nomen: No content designation in MARC now. Split authorities into two records? Names plus attributes of entity? Big undertaking Now in the authority format X51, 370,371, many fields with $z; location of meeting in X11
- Place: Already fields in place in both authority and bib formats
- New RDA guidance chapters
- Recording methods: For structured descriptions: Focused on manifestation statement, MARC 21is too granular, breaks up RDA element to pieces in 245 / series into subfields in 490 / No exactequivalence / How to fix? New indicators to existing fields or new subfields added to above ORnew field
- Representative Expressions: There are many elements of a representative expression, such asaspect ratio or medium of performance. A representative expression record would look similarto expressions, but there are no ways to explicitly state that something is a representativeexpression / some of the elements though move up to the work level by virtue of being in therepresentative expression / New indicators or subfields could help, though designating therecord as for the representative expression in the 075 / completely new tag. Current bib formathas places for things like aspect ration, color content, duration, script, sound content of item,and these could move into the authority format.
- Data provenance: things might be of interesting for provenance: agent who publishes metadata,transcription standard used, content standard used etc. Some of this in the authority and bibformats. Source for authorities are in 670 / missing elements for validity of metadata,transcription standard for metadata. Greater scope for content designation may be available inthe authority format.
- Diachronic works: Has a core concept of the PLAN, successive determinate or indeterminate;integrating determinate/indeterminate plans; static plan. Nothing in MARC21 for diachronicworks
- Timespan: accommodated in various places in the bib format, no current method for thetimespan element to be treated as a clump that can be pointed to outside of something like thesubject vocabularies
- Options for incorporating 3R changes are limited / BF, Schema.org could be extended to work,but there are no definite signs of development in response to LRM / 1-2 full MAC cycles toamend MARC ; Toolkit is still in Beta ; Could RDA/MARC Working Group reconvene? New PCCtask group? Other options?
Thoughts on the above? Invitation to comments from the room, CC:DA and the room.
- Is it worth developing MARC more to accommodate RDA? But RDF is also difficult for somethings as distribution and provenance change with each iteration. Hybrid records now areproblematic as far as metadata provenance, with mixed content standards. Should we freezewhere we are now? There are many half-solutions to in asking MARC to accommodategraph-like structures.
- Bib/Authority bifurcation is problematic. Bib record as stand-in for RDF graph…
- After all the comments no real consensus.
Would the room be interested in reconvening the RDA in MARC Working Group? Or should we gosomewhere else more forward-looking?
- If a group is formed it should include practitioners, as well as some vendor input. No immediateplan for directions.
- Timescale of changes? MAC’s process is deliberative, and a 1-2 cycle time plan is optimistic.Might MAC have a different cycle? And NACO node changes can take longer.
Subject Access Committee 6/25/18
Supporting Digital Humanities and LAM Data Access through Semantic Enrichment (Marcia L. Zheng, School of Information, Kent State University)
- Why digital humanities? It’s the intersection between the humanities and computing disciplines.Librarians in one study argued that 6/10 in a survey thought the work belongs in a library. At arecent meeting, a high proportion had librarians as presenters. Http://dev.diggingintodata.org isa program with much grant support, and has some interests in metadata work.
- Data doesn’t necessarily map into only digital data (think: tangible materials.)
- Data on the Web Best Practices (W3C 2017) encourage providing metadata.
- LAM world has many examples of structured data in things like archival finding aids, catalogs,curated research datasets. Often there will be unstructured portions of structured data (archivaldocumentation not in EAD, TEI files…) Unstructured data is what we have the most of, andcomes in the most varied forms, and are the most difficult to process.
- Data from LAMS and cultural heritage institutions are a rich source for humanities researchers.
- Trends in DH-LAM world: bigger data, more-structured data, machine-actionable data.
- Three perspectives on creating structured data:
- Documentation/metadata: Descriptive / administrative / structural/technical metadata
- Metadata: Descriptive metadata, indexing, markup, ontology Use Metadata, when Isomething cited researched tagged…
- An aside on the importance of provenance information: The famous Kent StateMassacre photo has been doctored to remove the pipe coming out of the head of thecentral figure. Going back through states of the image could help a user glean whatchanges had taken place, when.
- Semantic enrichment process
- Analysis: pre-enrichment phase
- Augmentation, selecting values from the contextual resource to enhance (See Europena Task Force on Enrichment and Evaluation report 29/10/2015)
- POSSIBLE APPROACHES
- An enrichment task: a process that improves metadata bout an object by adding newstatements “Enrichment” can mean several things: Alignment (Contextualize, createtyped relationships between resources of different types, usually in controlled form,align to sources like DBpedit, GEMET… ; “massage” the metadata, align with existingvocabularies to harvest external data, “sameAs” can lead to harvested content ; connectto “real thing,” e.g. Weissbib project, which links to external resources as part of theproject, added schema:sameAs, foam:focus, mapFAST project that can include a map.)
- “Expansion” has starting point, existing metadata components that are unstructured.Showed an example of a hybrid record with expansions of the original. Getting structured data from less structured parts of a description. Used OPEN CALAIS in one example to paste semi-structured text into the tool to extract entities in the text, entityextraction from EADs (persons, corporates, geographic, events are pretty accurate;social tags, industry terms and products categories are not trustworthy). These tools canbe good for description and identification, but not so good for analysis of content (e.g.for subjects). There are taxonomies and ontologies behind the tool. New structured datafrom unstructured data, one case looked at ETDs. Focus on abstracts, titles, keywordsand introductory paragraphs. Abstracts work better on titles, more useful for fields likemusic, where titles can be more fanciful. Major concepts were correct in most cases.
- Ontology design for non-structured data: Linked Jazz project used oral historytranscripts from 5 archives. Analyzed names mentioned to develop relationshipsbetween entities, e.g. mentors.
- Look at wiki.numismatics.org/numishare:visualize.
- How to deal with images: deep image/semantic analysis; look at annotations; semanticannotatioin goes deeper, enriches the unstructured or semi-structured tools.
- Tool: Mix’n’Match, lists entries of some external databases over 1000 catalogs andallows users to match against Wikidata items.
- Exposing our work: how do we get users to our sites?
- DPLA and places like WorldCat using Schema.org can extract much information. Thesecould be ways to get around bottlenecks in capacities to provide fuller subject access.
- In the end: If a vocab has embraced linked data we can extract much.