Linked Data: Why Bother?

7,669 views
7,534 views

Published on

Presented at the Northern Ohio Technical Services Librarians' meeting, November 22, 2013. Describes why libraries should move toward a linked data future to enable their resources to be discoverable on the open web, and includes lessons learned from developing the eXtensible Catalog at the University of Rochester.

Published in: Technology, Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
7,669
On SlideShare
0
From Embeds
0
Number of Embeds
5
Actions
Shares
0
Downloads
20
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Linked Data: Why Bother?

  1. 1. LINKED DATA: WHY BOTHER? JENNIFER BOWEN, UNIVERSITY OF ROCHESTER NOTSL MEETING, KENT STATE UNIVERSITY NOVEMBER 22, 2013
  2. 2. My Topics Today 2     The ―Vision‖ piece: Why should libraries care about linked data? A few linked data use cases for libraries Can libraries achieve their metadata-related goals WITHOUT linked data? Lessons learned from developing the eXtensible Catalog and what that has to do with linked data
  3. 3. eXtensibleCatalog.org 3
  4. 4. XC User Research Partners: Cornell University Ohio State University University of Rochester Yale University 4 Studying scholars at the UR…
  5. 5. Scholars want to read everything on the topic that they are researching 5
  6. 6. 6 They want to be in the middle of everything they need, all organized so it is findable and usable
  7. 7. Scholars want their research to be findable and usable by others. 7
  8. 8. “These other researchers cite MY research…” 8
  9. 9. Scholars want to connect to people whose work is interesting and useful to them. 9
  10. 10. Scholars don’t care what the technology is, as long as it helps them do their work 10
  11. 11. A shift in how people seek and use information 11    Systems that libraries provide (websites, catalogs, databases) are bypassed …not just in favor of Google and the Web in general …but also in favor of tailored desktop, mobile, and web applications
  12. 12. Beyond library finding tools 12 ―Even scholars who continue to use library finding tools are turning to new applications to aggregate and analyze information in ways that extend their scholarship beyond what manual searching and analyzing allows.‖ -- Nancy Fried Foster Senior Anthropologist, Ithaka S+R
  13. 13. Vision for how to address this… 13  Make library resources discoverable on the open web, through applications that potential readers are already using: Search engines Mobile apps Social media
  14. 14. AN EXAMPLE…
  15. 15. An example…Mt. Hope Cemetery 15 Photo credits: ROCHESTER’S SPEAKING STONES By Th. Emil Homerin; University of Rochester Department of Religion and Classics http://www.rochester.edu/College/REL/faculty/homerin/REL167/reports.htm
  16. 16. An example…Mt. Hope Cemetery 16 Photo credit: www.findagrav.com/cgibin/fg.cgi?page=pv&GRid=31&PIpi=76 016
  17. 17. 17 Photo credits: University of Rochester. River Campus Libraries. Department of Rare Books and Special Collections. http://www.lib.rochester.edu/index.cfm?PAGE=4119
  18. 18. What’s the role of linked data? 18   Tools like this are possible today with dedicated programming. Linked Data will enable library resources to be included in applications like this by allowing application developers access to a “…a store of machine-actionable data on which improved services can be built”. (Linked Open Data value statement)
  19. 19. THREE INITIATIVES RELATED TO LINKED DATA AND LIBRARIES
  20. 20. Stanford Linked Data Workshop (2011) 20 Linked Open Data Value Statements http://www.clir.org/pubs/reports/pub152/LinkedData Workshop.pdf
  21. 21. Linked Open Data Value Statements 21      Linked Open Data (LOD) puts information where people are looking for it: on the web LOD can expand discoverability of our content LOD opens opportunities for creative innovation in digital scholarship and participation LOD allows for open continuous improvement of data LOD creates a store of machine-actionable http://www.clir.org/pubs/reports/pub152/LinkedData data on which improved services can be built Workshop.pdf
  22. 22. More Linked Open Data Value Statements 22   Library LOD might facilitate the breakdown of the tyranny of domain silos LOD can provide direct access to data in ways that are not currently possible, and provides unanticipated benefits that will emerge later as the stores of LOD expand http://www.clir.org/pubs/reports/pub152/LinkedData Workshop.pdf
  23. 23. Another library linked data initiative: BIBFRAME 23 www.loc.gov/bibframe/
  24. 24. What is BIBFRAME? 24   Library of Congress-led effort to replace MARC 21 with a new bibliographic model based upon linked data ―Determine a transition path for the MARC 21 exchange format in order to reap the benefits of newer technology while preserving a robust data exchange that has supported resource sharing and cataloging cost savings in recent decades.‖
  25. 25. More goals of LC’s BIBFRAME 25    Differentiate between conceptual content and physical manifestations (works and instances) Focus on unambiguously identifying information entities (e.g. authorities) Leverage and expose relationships between and among entities http://bibframe.org/
  26. 26. 26 Potential Issues with BIBFRAME    Conceptual model doesn’t fully conform to either FRBR or RDA (e.g. no ―expression‖ level) – is this a problem? Will organizations that have already implemented linked data use BIBFRAME once it is finished? Do we really need a new serialization of MARC dictated by LC? http://bibframe.org/
  27. 27. Let’s get a little more specific… WHAT CAN LIBRARIES ACTUALLY DO WITH LINKED DATA?
  28. 28. 62 Use Cases for Library Linked Data! ―The mission of the Library Linked Data incubator group is to help increase global interoperability of library data on the Web, by bringing together people involved in Semantic Web activities—focusing on Linked Data—in the library community and beyond, building on existing initiatives, and identifying collaboration tracks for the future.‖ 28
  29. 29. W3C Library Linked Data (LLD) Incubator Group Use Case Areas 29         Bibliographic data Authority data Vocabulary alignment Archives and heterogeneous data Citations Digital objects Collections Social and new uses Source: Library Linked Data Use Cases
  30. 30. 30 Some Sample Use Cases W3C Library Linked Data Incubator Group: http://www.w3.org/2005/Incubator/lld/XGR-lld-usecase-20111025/
  31. 31. 31 Bibliographic Data Use Case: Deduplication and Unification of Library Records     Enable matching not based upon data from a central provider More reference data for matching/deduplication would be available openly for any library to use Non-MARC metadata also need deduplication and unification Using linked data would result in more trusted matches, more opportunities to automate the Source: Library matching process Linked Data Use Cases
  32. 32. Deduplication/Merging of Metadata With and Without Linked Data Using Linked Data 32 Record for Resource B: Match Point 1 Match Point 2 Deduping Records Record for Resource A: Match Point 2 Graph for Resource A URI: URI: URI: URI: URI: Graph for Resource B URI: URI: URI: URI: URI: If records match on a designated match point, one record overlays the other or a merge algorithm can keep data from both records Algorithm could look at all URIs representing two resources to determine a ―match‖ and combine all URIs into a single graph
  33. 33. Authority Data Use Case: Authority Data Enrichment (VIAF) 33     Enrich already existing authority data with additional information from external data sets by linking instead of copying & merging Enables VIAF (Virtual International Authority File) to be expanded with huge amounts of data from all over the world Align different representations of the same real-world resource Linked data allows the usage of remote data in applications Source: Library Linked Data Use Cases
  34. 34. Vocabulary Alignment Use Case: Vocabulary Merging 34    Users expect to be able to search for subjects using their own language and terms in an unambiguous, contextualized manner. Linked Data technologies could provide the underlying infrastructure by semantic mapping or merging of concepts across vocabularies. Allow vocabularies defined by different sources to organize (classify, index ...) legacy data to be used together Source: Library Linked Data Use Cases
  35. 35. 35 http://aep.lib.rochester.edu/hom e
  36. 36. 36
  37. 37. Keyword = ―arrow‖ 37
  38. 38. LCSH via id.loc.gov 38
  39. 39. Vocabulary Merging: Rochester AIDS Posters vs. LCSH 39 Arrow [URI for UR’s vocabulary AIDS poster terms] Same as Arrow (Symbol) id.loc.gov/authorities/ subjects/sh20130005 24
  40. 40. Archives and Heterogeneous Data Use Case: Semantic Connections 40   A group of archives would like to better share information about their holdings. They have separate catalogs and these catalogs do not necessarily use the same data formats. Exporting and sharing their data in Linked Data format would allow them to make connections between the collections using topics, names, place names, and other information contained in their metadata. Source: Library Linked Data Use Cases
  41. 41. 41
  42. 42. LCSH: AIDS (Disease)-Prevention 42
  43. 43. UR Local vocabulary: AIDS Prevention 43
  44. 44. 44 Rochester AIDS Posters vs. UCLA AIDS Posters: Semantic Connections AIDS prevention [URI for UR’s vocabulary AIDS prevention] Same as AIDS (Disease)— Prevention [URI for LCSH term]
  45. 45. Social and New Uses Use Case: Search Engine Optimization 45  Make library data searchable through Web search engines by:  Adopting an architecture that is compatible with web crawling by bots, and  Optimizing the available content so that search engines can process it efficiently  Adding structured metadata (e.g. RDFa) to library online catalogs could increase the visibility and accessibility of their data. Source: Library Linked Data Use Cases
  46. 46. 46 ―…the entire publicly available version of WorldCat is now available for use by intelligent Web crawlers, like Google and Bing, that can make use of this metadata in search indexes and other applications. ‖
  47. 47. LET’S TURN EVERYTHING ON ITS HEAD…
  48. 48. 48 Envisioning The Future Without Linked Data Or, What we learned from developing eXtensible Catalog (XC) software
  49. 49. What is XC software? 49 eXtensible Catalog (XC) is open source, user-centered, next generation software for libraries. XC provides a discovery system and a set of tools for libraries to manage metadata and build applications.
  50. 50. eXtensible Catalog Funders and Contributors 50 Major Funding Andrew W. Mellon Foundation Major Contributors Consortium of Academic and Research Libraries in Illinois (CARLI) Kyushu University University of Rochester
  51. 51. Why Did We Build XC? 51 Empower libraries to have control over their discovery environment Put results of user research into practice Extremely customizable user interface
  52. 52. Why Did We Build XC? 52     Create a new metadata management platform Implement a FRBR-based record structure Facilitate RDA implementation Repurpose MARC 21 records
  53. 53. ―FRBRized‖ MARC records 53 Parsing MARCXML records into linked FRBR-based XC Schema records XC Work Work Expressed XC Expression MARCXML Bibliographic ―Uplink‖= Record ID of the parent record created during OAIPMH harvest. Expression Manifested XC Manifestation
  54. 54. Facilitating RDA Implementation XC transforms MARC data into a FRBR-informed ―transitional‖ XML schema The ―XC Schema‖ uses a subset of RDA elements and roles alongside Dublin Core, some XC data elements More RDA elements can be added to the schema in the future 54
  55. 55. Repurposing MARC 21 records 55 Converts MARC codes to vocabulary values  Removes extraneous data  Normalizes inconsistencies  Maps most MARC fields/subfields and parse to appropriate FRBR Group 1 entity records 
  56. 56. 56 How XC Software Works (in a nutshell…)
  57. 57. How XC software works 57     Harvests a copy of metadata records in an existing repository Processes (cleans up, transforms) those records Makes records available for use in other applications Synchronize records in XC with records in original repositories …it’s all about metadata records!
  58. 58. eXtensible Catalog Architecture 58 Drupal MST OAI NCIP Toolkit Toolkit Toolkit Toolkit Metadata Services - Cleanup - Format Convert ILS Connectivity Synchronize data with XC User Interface - Search - Browse ILS Connectivity - Circ. status - Account info ILS ―Driver‖ Digital Repository ILS User Interface Metadata Live Circ. Data ILS ―Driver‖
  59. 59. eXtensible Catalog Architecture 59 Insert your Application with OAI-PMH Toolkit Harvester User Interface here! - Search Drupal - Browse MST OAI NCIP Toolkit Toolkit Toolkit Metadata Services - Cleanup - Format Convert ILS Connectivity Synchronize data with XC ILS Connectivity - Circ. status - Account info ILS ―Driver‖ Digital Repository ILS User Interface Metadata Live Circ. Data ILS ―Driver‖
  60. 60. 60 What we learned from ―FRBRizing‖ MARC in a live production system …three issues…
  61. 61. ―FRBRizing‖ MARC records 61 Parsing MARCXML records into linked FRBR-based XC Schema records XC Work Work Expressed XC Expression MARCXML Bibliographic ―Uplink‖= Record ID of the parent record created during OAIPMH harvest. Expression Manifested XC Manifestation
  62. 62. Linked Work, Expression and Manifestation Records in XC 62
  63. 63. 63
  64. 64. ―Uplinks‖ between FRBR levels 64
  65. 65. 65 Issue 1: Managing Relationships Parses MARCXML records into linked FRBR-based records How many FRBR entity relationships can we support with MARCXML Bibliographic XC software? ―Uplink‖= Record ID of the parent record created during OAIPMH harvest. XC Work XC Expression XC Manifestation
  66. 66. 66 Issue 1: Managing Relationships MARC bibliographic records can refer to multiple FRBR entities of the same type (analytics that represent multiple works/expressions, e.g. tracks on a CD) XC XC Work Work XC Work XC XC Expression Expression XC Expression MARCXML Bibliographic XC Manifestation
  67. 67. Issue 2: Beyond FRBR Group 1 Entities 67 MARC ―Alternate Graphic Representation‖ (880 fields) can contain data that belong in records for Group 2 and Group 3 entities Contributor: 700 1 ‡6 880-08 ‡a Vasil’ev, Maksim. 880 1 ‡6 700-08 ‡a Васильев, Максим. Subject: 600 10 ‡6 880-06 ‡a Putin, Vladimir Vladimirovich, ‡d 1952880 10 ‡6 600-06 ‡a Путин, Владимир Владимирович, ‡d 1952-
  68. 68. Issue 2: Beyond FRBR Group 1 Entities 68 If we were to parse this 880 data correctly, we would need to create and link to two additional records for Contributor and Subject that include the alternate scripts Contributor Subject (alternate forms from 880) •Contributor in Cyrillic characters •Contributor in Roman characters (alternate forms from 880) •Subject in Cyrillic characters •Subject in Roman characters XC Work XC Expression MARCXML Bibliographic XC Manifestation
  69. 69. 69 Issue 3: Related Group 1 Entities Language attribute for a related expression 041 1 ‡a eng ‡h ita 100 0 ‡a Dante Alighieri, ‡d 1265-1321. 240 10 ‡a Divina commedia. ‡l English 245 14 ‡a The divine comedy / ‡c Dante ; a new verse translation by C.H. Sisson. 500 ‡a Translation of: Divina commedia.
  70. 70. Managing Relationships 70 If we were to parse the original language from 041 ‡h, we would need to create and link to another ―based on‖ expression record (if we even have enough information to create it) Contributor Subject (alternate forms from 880) •Contributor in Cyrillic characters •Contributor in Roman characters (alternate forms from 880) •Subject in Cyrillic characters •Subject in Roman characters XC Work XC Expression Based on (Expression) – from 041 ‡h MARCXML Bibliographic XC Manifestation
  71. 71. 71 What XC has taught us about FRBR…  The GOOD news: MARC data is very rich, and contains data about MANY relationships described in FRBR and related data models There are hundreds of RDA Relationships between FRBR entitles!
  72. 72. What XC has taught us about FRBR 72 Maintaining links between separate FRBR entity records in a production environment is likely not scalable if we continue to manipulate records. XC Work •new records •changed records •deleted records •changed relationships XC Expression XC Manifestation
  73. 73. 73 What XC has taught us about FRBR…  The GOOD news: MARC data is very rich, and contains data about MANY relationships described in FRBR and related data models  The BAD news: managing all of these relationships in a record-based system is probably not feasible
  74. 74. RDA Implementation Scenario 1 (2007) 74
  75. 75. XC AND LINKED DATA: OUR ―AHA!‖ MOMENTS!‖
  76. 76. Our first ―Aha! Moment‖ 76  It would be much easier to ―FRBRize‖ MARC data using Linked Data than by creating and maintaining links between separate metadata records that have FRBR-related relationships to each other!
  77. 77. A Second ―Aha‖ Moment! 77 Creating Linked Data triples that refer to FRBR entities would be more meaningful than creating triples that refer to MARC records XC handles the interim step, of converting MARC data to FRBR entities
  78. 78. RDF triple 78 Subjec t This resource Predicat e Object has creator J. K. Rowling
  79. 79. With and without FRBR 79         Without FRBR: <MARCBibRecord-number> has_author ―J K Rowling‖ With FRBR: <Work-id> has_creator ―J K Rowling‖ <Expression-id> has_language ―English‖ <Expression-id> has_parent_work <Work-id> <Manifestation-id> has_isbn <ISBN-number> <Manifestation-id> has_parent_expression <Expression-id>
  80. 80. 80 Why use FRBR for Linked Data?    User research shows that users want to see the relationships between resources, etc. With XC, we can explore when/how FRBR might be useful for linked data Other data models may be more appropriate in some contexts and those can be explored as well.
  81. 81. Another not-quite ―AHA! Moment‖… 81 XC can serve as an interim step to create Linked Data because XC’s underlying schema uses elements from registered element sets (i.e. data elements already have URIs)
  82. 82. RDF Triple - Registered Data Elements 82 Subjec t oai:mst.rochester.edu: MST/ MARCToXCTransformatio n/ 10081 This resource Predicat e Object http://id.loc.gov/authoritie s/sh85103735#concept http://www. extensiblecatalog.in fo/Elements/subject has subject Poets, America n
  83. 83. XC Schema Properties 83 DC     Dublin Core terms (all) RDA – subset of elements and role designators XC elements (newly-defined) – when necessary All properties are from registered element sets and thus already have URIs RDA XC
  84. 84. 84 XC and Linked Data: What’s Next? XC facilitates associating metadata with FRBR Group 1 entities using data elements (mostly from RDA and Dublin Core) Implementing FRBR may help us create more meaningful Linked Data in some situations How can we make XC actually output Linked Data?
  85. 85. http://estc.bl.uk/ 85
  86. 86. http://estc21.wordpress.com / 86
  87. 87. eXtensible Catalog Architecture 87 Drupal Toolkit New ESTC User Interface Interface to be - Search built on Collex - Browse software MST OAI NCIP Toolkit Toolkit Toolkit Metadata Services Metadata Services - - Cleanup Cleanup - - Format Convert Format Convert ILS Connectivity Synchronize data with XC ILS Connectivity - Circ. status - Account info ILS ―Driver‖ Digital Repository ILS User Interface Metadata Live Circ. Data ILS ―Driver‖
  88. 88. ESTC Linked Data Benefits 88        Make data available for computational use Transform data back to MARC for reuse in library systems More granularity of data (e.g. date ranges) Collect new types of information, some not supported by MARC Incorporate information from other projects (VIAF) Make ESTC data more amenable to reuse by other projects, including discrete bits of data http://estc21.wordpress.com/data/
  89. 89. LINKED DATA CHALLENGES (WHY WE SHOULDN’T CREATE LINKED DATA?)
  90. 90. ―We won’t be able to control our data!‖ 90
  91. 91. Linked OPEN Data? 91      How much data to make available? Concerns about jeopardizing future business models Can we predict now how much data will be needed to fulfill future use cases? Metadata licensing issues Rights management
  92. 92. How will we assess quality? 92    Provenance: where did this data come from? Should ―triples‖ become ―quadruples‖ so we can tell ―who said this‖? Is the data accurate?
  93. 93. How can we maintain/improve quality? 93      How to manage data coming from multiple sources? What are best practices for improving it? Can we take advantage of information in application profiles? How/when should we aggregate metadata? What tools will we need?
  94. 94. 94 Next Steps: Continue the discussion!
  95. 95. Thank you! Additional photo credits: University of Rochester Photographic Services www.publicdomainpictures.net/view-image.php?image=54374&picture=runningbulls-12 www.dreamstime.com/stock-photos-group-kids-children-running-image5855523 www.publicdomainpictures.net/view-image.php?image=49200&picture=herd-ofhorses www.publicdomainpictures.net/view-image.php?image=42311&picture=oceanthrough-window-frame www.publicdomainpictures.net/view-image.php?image=10217&picture=goldenstar www.publicdomainpictures.net/view-image.php?image=27317&picture=hand-tools www.publicdomainpictures.net/view-image.php?image=27274&picture=stair-steps JENNIFER BOWEN JBOWEN@LIBRARY.ROCHESTER.ED U

×