Posting to RDA-L
On 02/03/2012 11:45, Bernhard Eversberg wrote:
Linked data originating from libraries have been made available some time ago, from several projects. Are there any reports as to their utility, their actual usage, and the quality and advantages of using them?
The only one I have any experience with (and that is very little) is the attempt by FAO of the UN to put the AGRIS database in RDF. AGRIS is a database that has been around since the 1960s, follows its own rules (non-ISBD), uses its own subject descriptors (AGROVOC) and is aimed primarily at agricultural technical experts. Here is an example of OpenAgris http://agris.fao.org/openagris/search.do?recordID=JP2010001379.
We can see how it automatically brings in statistics from another FAO database, and info from Wikipedia (through dbpedia). You can also see the other articles from the same journal issue (Fish Pathology). The rest of the record looks much as it always has, with title, author, description, keywords (from the AGROVOC descriptors). There is also an example of RDF (for another record) http://agris.fao.org/aos/records/AM2010000092, and the pure RDF/XML http://agris.fao.org/aos/records/data/AM2010000092?output=xml
I think this shows both the promise of Linked Data as well as the serious problems. For instance, when someone is looking at the first record above, what other information is useful for them to see? What use is it for a technical expert to see the Wikipedia entry on trout or salmon? The users of this catalog are experts after all, and the most anyone could hope for is that this expert would correct errors in Wikipedia, but that is not the idea. So, is this really useful from the point of view of an Agris user, or does it just take up space on the page? Also, while it is ok to see articles from the same journal issue, this is nothing new and has always been able to be done with authorized forms and consistent series numbering by a click (or cards brought together in the catalog). While a single journal issue may have articles on the same topic as the one I am looking at (e.g. a themed issue), there is certainly no guarantee and will probably have articles on different topics.
I am not finding fault here since, after all, you have to link to something and linking to dbpedia is probably better than linking to Pokepedia(!), but nevertheless, the ultimate utility for the user must be questioned. This is where true librarianship can make itself felt. What would be useful for the patron to see when looking at this record? What should they be looking at? This should be considered without being limited to what is in the linked data diagram.
I would say, at least Google Scholar, some proprietary databases, Google Books, Scirus, Bielefeld Academic, some blogs, and some specialized listservs. I am sure there are a lot more. Therefore, to create a really useful tool that provides relevant information that is in the patron’s information universe, these resources have not only to open up their “metadata” or their entire databases, the links of the linked data would have to be at least semi-coordinated. In my own opinion, getting this to work will be exponentially more difficult and complex than setting up the foundations of the Semantic web, which is important, but still a rather mechanical task. Our task will also be much more interesting. Somebody will do it sooner or later, I think, and I just hope libraries will be a part. What importance RDA or FRBR has to do with any of this is very, very slight.