RDF for Librarians
- 1. RDF FOR LIBRARIANS JENN RILEY METADATA LIBRARIAN DIGITAL LIBRARY PROGRAM DLP Brown Bag Series, September 22, 2010
- 2. From Wikipedia… “A collection of RDF statements intrinsically represents a labeled, directed multi-graph.” <http://en.wikipedia.org/wiki/Resource_Description_Framework> 9/22/10 2 DLP Brown Bag Series Learning the lingo is important; however for us it’s probably better to start with understanding how some of the basic concepts are different than what we’re used to. That’s what we’re going to do today.
- 3. Structural differences 9/22/10 3 DLP Brown Bag Series
- 5. RDF has “statements” and “graphs”5 Figures from RDF Primer <http://www.w3.org/TR/rdf-primer/> This is a “statement,” aka a “triple”. A statement is made in a particular direction. Statements combine to form graphs. A graph is of no fixed size and contains no predetermined types of statements. (The graph is the real RDF model; a triple is a secondary representation.) subject object predicate
- 6. More about statements 9/22/10DLP Brown Bag Series 6 subject: must be a URI object: can be a URI, or a “literal” predicate: must be a URI Figure from RDF Primer <http://www.w3.org/TR/rdf-primer/>
- 7. The centrality of URIs in RDF 7 RDF uses URIs to identify: individuals, e.g., Eric Miller, identified by http://www.w3.org/People/EM/contact#me kinds of things, e.g., Person, identified by http://www.w3.org/2000/10/swap/pim/contact#P erson properties of those things, e.g., mailbox, identified by http://www.w3.org/2000/10/swap/pim/contact# mailbox values of those properties, e.g. mailto:em@w3.org as the value of the mailbox property (RDF also uses character strings such as "Eric Miller", and values from other datatypes such as integers and dates, as the values of properties) Figure and text from RDF Primer <http://www.w3.org/TR/rdf-primer/>
- 8. Implications of statement structure (1) 9/22/10DLP Brown Bag Series 8 Subjects are always formalized; know exactly what is being talked about They’re only implicit in library metadata Makes moot the 1:1 problem Might still have content vs. carrier problem Predicates are always formalized Maybe not all that different than library/archive/museum metadata models More obviously reusable
- 9. Implications of statement structure (2) 9/22/10DLP Brown Bag Series 9 Representation of objects is flexible Using a URI facilitates connecting this value into a larger graph Literals allow more limited functionality; used when the value isn’t likely to be useful as part of a vocabulary URI vs. literal a design choice; note the previous example used a literal where a URI might be better! Making a statement that two different URIs refer to the same thing can join two graphs Note, URIs aren’t necessarily dereferencable
- 10. Blank nodes 9/22/10DLP Brown Bag Series 10 Figure from RDF Primer <http://www.w3.org/TR/rdf-primer/> This blank node stands in for “John Smith’s address”
- 11. Properties vs. elements 9/22/10DLP Brown Bag Series 11 Remember, a statement’s predicate represents a property of the resource (subject) being described Resource=subject; Property=predicate. I don’t know why there are different terms for the model than for individual statements. Let’s move on. On the surface, an RDF property looks like the same thing as an XML element or a database field But the underlying formal model is different Explicit subjects and directionality of statements Formality of RDF model places additional restrictions but allows more explicit meaning Therefore inferences you can draw for elements and fields are not as strong as you can for properties
- 12. Graph vs. tree 9/22/10DLP Brown Bag Series 12 Tree vs. graph figure from XML2RDF documentation XML documents are trees This is as good as you can do inferring a graph from a tree:
- 13. Model vs. its syntax(es) 9/22/10DLP Brown Bag Series 13 There’s a difference between: Model of information representation Property or element definitions Binding of the information into a specific syntax MARC is both the record structure (syntax) and content designations (element definitions) RDF model has many encoding syntaxes RDF and MARC operate at different levels in this landscape; but that doesn’t mean MARC’s structure is capable of representing the RDF model
- 14. Some realities of RDF that scare us 9/22/10DLP Brown Bag Series 14 No predetermined set of properties to care about No guarantee that the same person/item/place/whatever are always referred to with the same URI No inherent mechanism/requirement for vetting properties, URIs, etc But let’s be frank here. Are our library/archive/museum records really: Complete? Authoritative? Consistent? Accurate? All that functional for what we want to do?
- 15. More RDF concepts 9/22/10DLP Brown Bag Series 15 Class A statement can say a subject is of a certain type or category known as a “class”. Classes can be formally defined in RDF Schema documents Domain and Range Specify what classes subjects and objects of statements (respectively) using a given property can be More than one domain or range statement can be made for any given property These features allow “inferencing” to deduce statements that aren’t actually made
- 16. But both have… 9/22/10DLP Brown Bag Series 16 Raging debate over how precise you have to be in indicating what it is you’re describing An interest in mechanisms to allow consumers of data to evaluate the authority of statements A tendency to over-model (see “The Modeller” blog post on handout) Other similarities “Application profile” notion applicable to both XML and RDF models, though would be implemented differently XML elements repeatable; RDF makes no restriction on how many statements about the same resource use the same predicate.
- 17. Terminology differences 9/22/10 17 DLP Brown Bag Series
- 18. “Subject” 9/22/10DLP Brown Bag Series 18 In libraries, what an information resource is about In RDF, what a statement is about
- 19. “Vocabulary” 9/22/10DLP Brown Bag Series 19 In libraries, implies a controlled vocabulary of a certain sort Authorized terms Lead-in terms (see references, etc.) Sometimes, hierarchical structure Sometimes, related terms In RDF, much looser definition Includes formal definitions of classes and properties
- 20. “Class” 9/22/10DLP Brown Bag Series 20 In libraries, a classification scheme indicating the general topic or area of knowledge covered by an information resource In RDF, a type or category that any type of object or resource belongs to
- 21. “Schema” 9/22/10DLP Brown Bag Series 21 XML Schema defines a set of elements intended to be used together RDF Schema defines classes and properties intended to be used anywhere, alone or in combination
- 22. So, wait, why should I learn this? 9/22/10 22 DLP Brown Bag Series
- 23. Libraries are moving in this direction 9/22/10DLP Brown Bag Series 23 At least a little bit Increasing viability and acceptance of interoperating with data from outside of libraries RDA gives us an opportunity to fundamentally rethink some features of our data Semantic Web activities have been given new life with the grassroots Linked Data movement
- 24. Linked Data design issues 9/22/10DLP Brown Bag Series 24 Use URIs as names for things Use HTTP URIs so that people can look up those names. When someone looks up a URI, provide useful information, using the standards (RDF, SPARQL) Include links to other URIs. so that they can discover more things. Tim Berners-Lee <http://www.w3.org/DesignIssues/LinkedData.html>
- 25. How much planning do we need to do? 9/22/10DLP Brown Bag Series 25 “There are other cases where the easiest thing for somebody to do is to just put data up in whatever form it's available.” Tim Berners-Lee <http://www.readwriteweb.com/archives/interview_with_ti m_berners-lee_part_1.php> Let’s review some of what libraries have done so far.
- 26. id.loc.gov 26
- 29. LIBRIS – Swedish National Library29
- 30. Open Library 30
- 31. Registering FRBR and RDA properties31
- 32. What about incorporating data from other sources? 32 BBC Wildlife Finder <http://www.bbc.co.uk/wildlifefinder/> Slide by Thomas Baker, “What Makes the Linked Data Approach Different,” NISO DCMI Webinar 2010. <http://dublincore.org/resources/training/NISO_Webina r_20100825/dcmi-webinar-02.pdf>
- 33. Challenges to implementing RDF in practice 9/22/10 33 DLP Brown Bag Series
- 34. How do you… 9/22/10DLP Brown Bag Series 34 find all the triples you need? know what predicates and URIs to use when creating new triples? know what predicates and URIs to use when processing data?
- 35. What infrastructure needs to be built in order to… 9/22/10DLP Brown Bag Series 35 negate an erroneous statement? say that a statement is time-dependent? judge the likely validity of a statement made by someone else? …actually end up with machine-understandable data in the end???
- 36. Don’t run away! 9/22/10DLP Brown Bag Series 36 These uncertainties do not give librarians permission to write off the entire model. Incorporating new ideas doesn’t mean we have to give up our core principles. We have an opportunity here: to create classes and properties that represent data as we think it should be. to bring new features to the model and to its implementations. to provide some of the infrastructure that’s missing. to participate in the discussion.
- 37. Let’s go! 37 Photo by Gregory Moine <http://www.flickr.com/photos/gregory- moine/4713276677/> Creative Commons Attribution 2.0 Generic
- 38. For more information 9/22/10DLP Brown Bag Series 38 jenlrile@indiana.edu These presentation slides: <http://www.dlib.indiana.edu/~jenlrile/presentations/bbfall10/r df/rdfForLibrarians.pptx> Recording of talk on DLP Brown Bag page: <http://www.dlib.indiana.edu/education/brownbags/> More resources on handout: <http://www.dlib.indiana.edu/~jenlrile/presentations/bbfall10/r df/handout.pdf> Thank you!