Dewey Linked Data: An interview with Michael Panzer
Andy Havens, Coop Blog Editor: Michael... thanks for taking the time to talk to us about Dewey linked data. That's some cool stuff you're doing there.
Michael Panzer, Assistant Editor DDC: Yes, yes it is! We've been excited for awhile now about getting the full Dewey out as linked data.
Andy: The original experimental project started... what... about two-and-a-half years ago?
Michael: Closer to three years, but yes. We originally had linked data in the dewey.info experiment for around 1,100 DDC Summaries and after a year or so added 4,500 more classes per language by releasing the Abridged Edition in three languages.
Andy: And now there's linked data for all the DDC 23 classes.
Andy: Why the gap in time between the original experiment and the full implementation?
Michael: Well, first of all we were working to actually get DDC 23 out the door, so that had to take precedence. But also, linked data is a relatively new field. We had some ideas going into the project, but really wanted to work with the community to figure out what would be the most useful, stable and helpful way to represent Dewey information in a linked data model.
Andy: So the plan had always been to do the whole thing?
Michael: Yes, very much so. Even at the start, we wanted to eventually provide URIs for every Dewey class. But there were lots of questions to be answered and we needed to get some experience for ourselves, of course. Between the initial phase and today, we were able to clean things up quite a lot and make some changes based on excellent feedback from librarians and linked data experts.
Andy: Without getting too technical, can you tell us a bit about the linked data model you used?
Michael: We built our model based on SKOS (Simple Knowledge Organization System). SKOS is used as a Semantic Web data model for knowledge organization systems like thesauri. It's a good match with Dewey, even if it has some shortcomings accommodating a large universal classification system like the DDC. But SKOS is emerging as the smallest common denominator for knowledge organization systems on the web.
Andy: So that's the linked data vocabulary that you're matching up to Dewey. Is that similar in some ways to the addition of Schema.org linked data to WorldCat.org that was announced last month?
Michael: Yes, exactly. In order to make our data more useful to more developers outside the library world, it's helpful to use a linked data vocabulary that can help match our information up with what others are doing. This is also the reason why we didn't want to extend SKOS it too much in order to make it easy for people not familiar with Dewey to consume DDC linked data.
Andy: Can you give me an example of how someone might mash-up Dewey linked data with other data sets?
Michael: Well, geographic data comes to mind first. It might be very interesting to create a service that took Dewey classes and geographic information - which is already available as linked data - and put them together to do something like searching for resources about a subject based on where it was published or where the author was born, etc. Really, any time when you'd like to "overlap" information from different disciplines to see the intersection, linked data is going to be very useful. Over time, we expect to see lots of interesting apps and mash-ups of Dewey data that make it easier for librarians and end users to examine resources in new, interesting and, I think, fun ways. Just think about the potential for knowledge discovery if we could create a robust way for a user to scope a topical search to geographic area and time period!
Andy: That sounds really interesting. It's like being able to define entirely new facets for searches.
Michael: Exactly. And we hope that people will see Dewey as a great way to help people browse and locate materials by topic. Also, since Dewey numbers are concept-based, but not language specific, it will be much easier to build multi-lingual or language-agnostic apps. At the moment, the new DDC linked data is only in English - partly because DDC 23 is still relatively new - but adding more languages will be relatively easy and won't change the URIs currently in place.
Andy: I'm looking forward to seeing those results. So... what's next for Dewey linked data?
Michael: Well, we'll be adding the geographic Table 2 data, which would make it even easier to match up with GeoNames data. We're also looking to expose the Dewey data as links to FAST, which would allow for a kind of real-time crosswalk between Dewey and LCSH (Library of Congress Subject Headings).
Andy: That's one of the strengths of linked data, yes? Being able to help match concepts and terms from different services or ontologies?
Michael: Yes. With VIAF being an excellent example of that.
Andy: Anything else you'd like to add before we wrap up?
Michael: Two quick things. In terms of licensing, all this linked data are reusable under the same terms as before(Creative Commons BY-NC-ND) and the license is carried in each record. Also, dewey.info is now fully integrated in our standard workflow for the Dewey ecosystem. So any updates to the DDC will be reflected in the linked data.
Andy: Good to know. Thanks for taking the time to talk with us today.
Michael: Don't mention it.
Listed below are links to blogs that reference this entry: Dewey Linked Data: An interview with Michael Panzer.
TrackBack URL for this entry: