Enhancing a library OPAC with linked data
•Download as PPTX, PDF•
1 like•1,794 views
Experiments with semantic web markup and linked data for libraries. Loading and utilizing URI's on library MARC catalog records. Leveraging id.loc.gov name authorities links to connect patrons to WorldCat Identities.
Recommended
More Related Content
What's hot
What's hot (20)
Similar to Enhancing a library OPAC with linked data
Similar to Enhancing a library OPAC with linked data (20)
More from Michael Cummings
Recently uploaded
Recently uploaded (20)
Enhancing a library OPAC with linked data
- 1. Enhancing an OPAC using linked data in MARC records ELUNA 2016 Oklahoma City
- 2. ExLibris Users of North America Oklahoma City, OK May 2016 Presented by Michael Cummings Library Systems Coordinator GWU Libraries
- 4. HTML markup in web pages The semantic web
- 5. Google Trends Trend: ‘linked data’ vs ‘semantic web’
- 7. HTML markup in web pages In practice: an example findit.library.gwu.edu/isbn/9780716703389
- 8. Schema.org microdata markup Markup, generated on-the-fly
- 9. Google search Catalog record in Google search result
- 10. Custom Google Search form Custom Google search form and results reports.library.gwu.edu/customsearch.html new horizons in astronomy
- 11. Creating the custom search Experiment: Build a Google search form https://cse.google.com
- 12. Custom search coding ?????????????? Embedding the custom search form view-source:reports.library.gwu.edu/customsearch.html
- 13. 2. Hyperlinks on MARC data
- 14. Voyager-l informal survey Informal survey of Voyager-l
- 15. Voyager-l informal survey Informal survey of Voyager-l
- 16. About our web applicaton Beyond semantic markup
- 17. OCLC linked data: Identities Brandt, John C. Publication timeline, list of works, related names and subjects Use case: add value: WorldCat Identities findit.library.gwu.edu/oclc/327452
- 18. OCLC Identity example Overview, Timeline, Alternate names … 1 of 4 www.worldcat.org/wcidentities/lccn-n50012763
- 19. OCLC Identity example Works by or about the author … 2 of 4
- 20. OCLC Identity example Audience, Related Identities, Links … 3 of 4
- 21. OCLC Identity example … and Associated Subjects word cloud. 4 of 4
- 22. Identities links in the web application
- 23. Links on our MARC records STEP #1 Add URIs to local MARC records findit.library.gwu.edu/item/2659054.json “0”: http://id.loc.gov/authorities/names/n82076128” “o”: http://worldcat.org/entity/work/id/1419206”
- 24. Converting links STEP #2 convert name to ID link See: https://github.com/gwu-libraries
- 25. Publishing links in our API #3 Output link text and associated URI See: http://findit.library.gwu.edu/api/
- 26. Adding links with MarcEdit
- 27. Catalog record before editing Original local Voyager catalog record
- 28. Adding links with MarcEdit New in MarcEdit: MarcNext See: http://marcedit.reeset.net/
- 29. MarcEdit build linked records MarcNext: Link Identifiers: Build links
- 30. Links generated by MarcEdit Revised local Voyager Marc record $a Levin, Murray Burton. $0 http://id.loc.gov/authorities/names/n80034895 $0 http://viaf.org/viaf/170682523
- 31. OCLC Workid Example: URI for OCLC’s Work ID $n OCLC Work Id $o http://worldcat.org/entity/work/id/62586849
- 32. URIs on MARC tags in Voyager Count of URIs on GW’s local Marc records https://github.com/gwu-libraries/notebooks/blob/master/20151026-marc-subjects-spark/subject-patterns-with-spark.ipynb
- 33. Two experiments with linked data
- 34. Experiment: Display the links #1 Links displayed with labels . . .
- 35. Experiment: Display the links Cancelled before going into live site.
- 36. Experiment: Display the links #2 Use case: get alternate names 1 of 3
- 37. Experiment: Display the links and use labels from the model . . . 2 of 3
- 38. Experiment: Display the links . . . however, how fast can we process it? 3 of 3
- 39. Ex Libris and linked data ExLibris support for linked data
- 40. Ex Libris prototype going in the wrong direction.
- 41. Bibframe / data model
- 42. Bibframe: Library of Congress A model for expressing & connecting bibliographic data www.loc.gov/bibframe/
- 43. Zepheira, NLM, UC-Davis, GW Testing bibframe with library data bibfra.me/about/
- 44. RDF Bibframe version of MARC record bibframe.org/tools/compare/bibid/17930908
- 45. OCLC Turtle | N-Triple | JSON-LD | RDF/XML Available formats / APIs include RDF
- 46. Future Bibframe / RDF developments
- 47. Going forward: obstacles Experimentation and evaluation https://ecommons.cornell.edu/handle/1813/42435
- 48. Going forward: modelling Creating linked data repositories
- 49. Going forward: programming Open source tools to try out. pybibframe https://github.com/zepheira/pybibframe Converting MARC/XML to RDF or Versa output (command line) or Converting MARC/XML to RDF or Versa output (API)
- 50. Learn More Michael Cummings Library Systems Coordinator 202-994-4806 michaelc@gwu.edu Miller, E., & Ogbuji, U. (2015). Linked data design for the visible library. Bulletin of the American Society for Information Science and Technology (Online), 41(4), 23-29. Retrieved from http://proxygw.wrlc.org/login?url=http://search.proquest.com/docview/1679220450? accountid=11243 Mitchell, E. T., PhD. (2016). The current state of linked data in libraries, archives, and museums. Library Technology Reports, 52(1), 5-16,2. Retrieved from http://proxygw.wrlc.org/login?url=http://search.proquest.com/docview/1752517266?accou ntid=11243 Shieh, J. (2013, Winter). A transformative opportunity: BIBFRAME at the George Washington University, an early experimenter. Information Standards Quarterly,25, 17-21. Retrieved from http://proxygw.wrlc.org/login?url=http://search.proquest.com/docview/1545644349?accountid=11243 The Relationship between BIBFRAME and OCLC’s Linked-Data Model of Bibliographic Description: A Working Paper http://www.oclc.org/content/dam/research/publications/library/2013/2013-05.pdf
- 51. Photo credits Michael Cummings Library Systems Coordinator 202-994-4806 michaelc@gwu.edu Slide 1: Buffalo sketch https://www.pinterest.com/pin/390194755188382990/ Juan Bosco / San Martin Arts Crafts Slide 4: Oklahoma state bird: Scissor-tailed flycatcher in flight http://www.birdsasart-blog.com/2012/07/13/you-be-the-judge-flight/ Slide 13: Lone buffalo in tall grass https://upload.wikimedia.org/wikipedia/commons/thumb /e/e7/Buffalo_American_animal.jpg/1024px-Buffalo_American_animal.jpg Slide 22: Scenic turnout, near Arbuckle Mountains. http://www.travelok.com/scenic_drives Bill Bryant Slide 26: Indian Paintbrush wildflowers near Boynton. http://www.travelok.com/scenic_drives Slide 33: Oklahoma's state wildflower, the Indian Blanket, near Davis. http://www.travelok.com/scenic_drives Carolyn Fletcher Slide 41: Group of bison on grassy plain http://www.nature.org/cs/groups/webcontent/@web/@oklahoma/ documents/media/tallgrass-prairie-bison-2.jpg Slide 46: Two bison in snow Brett French, Billings Gazette
Editor's Notes
- Thanks to The University of Central Oklahoma, Oklahoma State University, and University of Oklahoma at Oklahoma City for hosting ELUNA 2016.
- For those who may not be familiar with George Washington University, we are a member of a consortium of nine academic libraries with over 12 million bib records in Voyager. (American, Catholic, University of DC, Gallaudet, George Mason, George Washington, Georgetown, Howard, Marymount)
- This presentation is divided into three main topics. The material is geared towards developers, but catalogers and administrators with an interest in linked data should find this information useful.
- The gist of the “semantic web” is that the HTML markup of web pages should be more than encoding for the browser. It should be encoded in a way that facilitates sharing or reuse. You might think of for example, tagging this scissor-tailed flycatcher by its domain, kingdom, phyla, class, order, family, genus, and species
- I recently found that the term ‘linked data’ is searched more often than ‘semantic web’ -- as shown in this Google Trends timeline. Linked data seems to have overtaken semantic web around 2012. But let’s look at semantic web anyway.
- The semantic markup vocabulary for web pages proposed by schema.org was an early development sponsored by Google, Microsoft, Yahoo and Yandex. Never heard of ‘Yandex’? It’s the top search engine in Russia.
- At GW we developed a custom web application as the front end to our consortium’s Voyager catalog. We incorporate Schema.org markup in the application-generated web pages. Let’s look at the HTML for this page...
- You can see this page tagged as type ‘Book’, with item properties cover-image, name, author, contributor, publisher (and more). Again, this encoding is NOT on the catalog record, it’s generated on the fly when the page is requested.
- We have a site map on our server -- basically a URL for every bib record -- which search engines crawl and index. Consequently, a web search may include the library record as in this example. (It helps to add ‘gw libraries’ as part of the search terms”) .
- A customized Google search form is also possible, like this example. I used Chrome a plug-in, AdBlock, to hide promoted results. This only shows results from our domain. I’m not sure what percent of the 12 million bib records have been indexed, but it’s interesting all the same.
- By the way, if you would like to experiment with a custom google search form, you can easily build a “custom search engine” or “CSE” at cse.google.com by completing a profile like this.
- The custom search engine will generate a javascript function and tag which you can copy and paste into your HTML page.
- That’s all I wanted to cover about semantic web markup for today. Let’s move on to the second part of the presentation, which is adding hyperlinks on MARC data.
- I did an informal survey of members of the Voyager-l listserv about hyperlinks from library OPACs. Other than Google Books Preview, there were very few external links identified.
- No one in my small response group is using a link to OCLC WorldCat identities, and only a few were familiar with the WorldCat Identities. So I’ll tell you how we came to add that link to our OPAC.
- [1] I mentioned we have a consortium catalog. GW’s custom catalog can get availability from two external ILS, and includes links to ILL and consortium loans; [2] the app uses APIs from Summon, 360 Link and other resources [3] and now handles URIs, specifically links to WorldCat Identities.
- We believe providing our patrons with links from the catalog details out to WorldCat Identities ADDS VALUE. This is how the link appears on the page. Let’s follow the link to see what we might find, for a different author.
- A long page of information is provided about the author. The first part is an overview. There is a publication timeline showing when there were publications by or about the author, and a list of alternate names is provided in the right-hand sidebar.
- Scrolling down we have a list of most widely held works by or works about the author, more alternate names, a facet for languages, and sometimes, book covers.
- There is an rating for ‘audience level’, a section for related identities, and ‘useful’ links … (actually I don’t find the audience level very useful).
- … and then finally at the end of the page we find a word cloud of associated subject terms. Of course the user could follow any of the links for even more information. Pretty cool.
- Next, let’s look at how we made the link from the OPAC to WorldCat Identities
- First of all, we need a URI link. We store them on our local MARC record in the subfield zero, or in subfield o for the 787 tag.
- The Identities link requires a step to convert the author name URI, preferably from id.loc.gov to the URL for the WorldCat Identities. This is a python example, which does a few other steps as well.
- Here is the result, the original URI from id.loc.gov for the author, and the WorldCat Identities link which is not stored on the MARC record. This screenshot shows a snippet of the JSON view from our web application.
- Next we look at how we add the URIs to the MARC records.
- Here is a local catalog record. Notice it does not have any URIs on MARC tags 100, or 6xx. OK, I’m going to save it as a ‘.bib’ file on my workstation.
- Now I’ll start MarcEdit. There is a relatively new sub-section of MarcEdit called MarcNext. It has tools BibFrame Testbed, JSON Object Viewer, SPARQL Browser, and Link Identifiers.
- Now we run Build Linked Records which has a number of options as you see here. Then I Process the build . . .
- … and I get a revised MARC record which I can load into Voyager. Notice it has several URIs now.
- As an aside, note that 787 ‘Other Relationship Entity’ uses subfield letter ‘o’ whereas all other URI’s are using subfield number 0 Now if we could index the 787 work ID value, we could programmatically create a collection of records having the same work ID.
- We did a lot of batch processing of our records. The URI count as of October, 2015. Total links: four million As another aside, developers might want to checkout the python notebook regarding a programmatic way of analyzing the URI distribution.
- Next, I would like to summarize two experiments using linked data related to library records. (Oklahoma state flower Indian Blanket)
- Earlier I showed you that we added a link to WorldCat Identities. On our development site we experimented with a “References” tab, where we listed list WorldCat Identities, WorldCat Works…
- …and then we listed ‘preferred terms’ and links to the authorities. We concluded there were too many distracting link; we didn’t think the end-user would find added value looking at most of the target pages.
- We experimented with the idea that the external sites could give our users help transliterating non-roman names. We mocked up some pages showing how we might get data from external links as JSON or other formats, then parsing the result and displaying information in our interface.
- …this experiment also presented different labels, like ‘Concepts’…
- Although we could retrieve the data, parse it, and incorporate it into our interface we are concerned about the timing of getting the data, and the complexity of parsing the data. Take alternate names for example, in some cases there may be just too many. (look at Mark Twain lccn-n79021164/ for example)
- ExLibris has published information regarding the company’s plans to support linked data in libraries. There are several ELUNA sessions that are related to the topic.
- However, based on this prototype screen in the white paper, I’m afraid they’re going in the wrong direction. Patrons don’t get much value, if any, from a list like this example. I hope the committees working on this topic get it right.
- Let’s wrap-up with a walk-through on the bibliographic framework model, or ‘Bibframe’
- Bibframe, according to the Library of Congress and partners, is a replacement for MARC, a general model for expressing and connecting bibliographic data.
- GW is one of the institutions collaborating with others on testing Bibframe models.
- There are tools to express MARC records in the Bibframe model… At GW we’re experimenting with exporting MARC to an RDF datastore with a BlackLight / Solr discovery interface.
- … earlier I mentioned retrieving WorldCat information using API’s. You can get data from WorldCat in the RDF format used by bibframe…
- There seems to be potential in Bibframe, but it is relatively new, and there are pros and cons.
- Cons: There are many obstacles to adoption. Lack of ILS support, discovery, training…this is a good article.
- Pros: Zepheira is working with several academic libraries… note, we’re talking about transforming the collection into linked data, and sharing it / publishing it on the web.
- On the plus side, there is a Python library, pybibframe, for converting MARC/XML to RDF.
- Well, that about wraps up the overview of our linked data experimentation. This is an interesting topic, so I’ve added some links for you to check out when you have time. As you know, these slides will be available online.
- Thanks for listening. I hope you also enjoyed some of the Oklahoma themed photos!