For our fed search service we very much echo Jonathan's real-time requirements/use case (we don't build indexes, so bulk download is not of interest):
access - real-time query (purpose - to enhance data about items found by other means) query - by standard IDs (generally this is "known item" augmentation, so "discovery" queries by keywords, etc are not so much required) data format - almost anything "standard" (we can translate it into the internal data model structure) big value add - relationships, mainly the "upward" ones, towards work data quantity - all details of directly related items, plus 2nd level links, possibly all details all the way up to (and including) the work (this is a trade-off of processing time on the service side to gather this information, and on our side to de-construct vs. the time to set up and manage multiple service calls to get the data about individual items in the link chain. In our experience it is almost always quicker to get it "all-at-once" than to send repeated messages, even if the total amount of data is less in the latter. But, mileage may vary here.) Peter > -----Original Message----- > From: Code for Libraries [mailto:code4...@listserv.nd.edu] On Behalf Of > Jonathan Rochkind > Sent: Wednesday, April 21, 2010 7:59 AM > To: CODE4LIB@LISTSERV.ND.EDU > Subject: Re: [CODE4LIB] What do you want out of a frbrized data web > service? > > So, okay, the "value added" stuff you have will indeed be relationships > between entities, which is not too unexpected. > > So, yes, I would want a real-time query service (for enhancement of > individual items on display in my system) _as well as_ a bulk download > (for enhancing my data on indexing). > > For real time query, I'd have a specific entity at hand roughly > corresponding to a 'manifestation'. I'd want to look it up in your > system by any identifiers I have (oclcnum, lccn, isbn, issn; any other > music-related identifiers that are useful?) to find a match. Then I'd > want to find out it's workset ID (or possibly expression ID?) in your > system, and be able to find all the OTHER manifestations/expressions in > those sets, from your system, with citation details about those items. > (Author, title, publisher, year, etc; also oclcnum/lccn/isbn/issn/etc > if > available. Just giving me Marc with everything might be sufficient). > If you have work identifiers from other systems that correspond to your > workID (OCLC workID? etc), I'd want to know those. > > For bulk download, yeah, I'd just want everything you could give me, > really. > > Some of the details can't really be spec'd in advance, it requires an > interative process of people trying to use it and seeing what they need. > I know this makes things hard from a grant-funded project management > perspective. > > Jonathan > > Riley, Jenn wrote: > > On 4/20/10 7:18 PM, "Jonathan Rochkind" <rochk...@jhu.edu> wrote: > > > > > >> But first, to really answer the question, we need some more > information > >> from you. What data do you actually have of value? Just saying "we > have > >> FRBRized data" doesn't really tell me, "FRBRized data" can be almost > >> anything, really. Can you tell us more about what value you think > >> you've added to your data as a result of your "FRBRization"? What > do > >> you have that wasn't there before? Better relationships between > >> manifestations? Something else? > >> > > > > Heh, I was intentionally vague in an attempt to avoid skewing the > discussion > > in certain directions, but I was obviously *too* vague - my apologies. > Here > > are the sorts of things we'd imagined and are looking to prioritize: > > > > - Give me a list of all manifestations that match some arbitrary > query terms > > - Given this manifestation identifier, show me all expressions on it > and > > what works they realize > > - Give me a list of all works that match some arbitrary query terms > > - Given this work identifier, show all expressions and manifestations > of it > > - Show me all of the people who match some arbitrary query terms > (women > > composers in Vienna in the 1860s, for example) > > - Which works have expressions with this specific relationship to > this > > particular known person? > > > > Basically we're exploring when we should support queries as words vs. > > previously-known identifiers, when a response will all be a set of > records > > for the same entity vs. several different ones with the relationships > > between them recorded, to what degree answering a query will involve > > traversing lots of relationships - stuff like that. Having some real > use > > cases will help us decide what kind of a service to offer and what > > technology we'll use to implement that service. > > > > We do hope to also be able to publish Linked Data in some form - > that's > > probably going to come a little later, but it's definitely on "the > list". > > > > To answer one of your other questions, the V/FRBR project is focusing > on > > musical materials (scores and recordings) in particular, but we hope > to set > > up frameworks that would be useful for library bibliographic and > authority > > data in general. > > > > Jenn > > > > ======================== > > Jenn Riley > > Metadata Librarian > > Digital Library Program > > Indiana University - Bloomington > > Wells Library W501 > > (812) 856-5759 > > www.dlib.indiana.edu > > > > Inquiring Librarian blog: www.inquiringlibrarian.blogspot.com > > > > > >
