-----Original Message----- From: Resource Description and Access / Resource Description and Access [mailto:[EMAIL PROTECTED] On Behalf Of Rhonda Marker Sent: Thursday, October 23, 2008 4:29 PM To: RDA-L@INFOSERV.NLC-BNC.CA Subject: Re: [RDA-L] FRBR user tasks (was: Alternatives to AACR2/MARC21?)
I've been lurking since the list began, but will dart out into the open just this once. You ask what difference it will make to try to bring both precision and recall to searches (my vocabulary, not yours). For some tasks, such as finding enough to write an undergraduate essay, perhaps itadds very little beyond what a simple keyword search would accomplish. For other tasks, a more comprehensive result is vitally important-- for graduate level research, for many STM (science-technology-medicine) topics, and I'm sure others could give many more categories of things and even specific instances. I guess my point is that the General Searcher is not our only user. Having a model like FRBR helps us organize our efforts so that whatever our resources allow us to do, we can do in a purposeful, concerted way. --------------------------- I haven't read that FRBR or RDA will make it easier for people to find things, except in the sense that the FRBR displays may provide more useful collocation of similar records (by work/expression/manifestation/item). And when we bring "precision and recall" into the equation, I don't know if this has anything to do with RDA or even with human cataloging. Traditionally, precision and recall have had to do with evaluating the results of automated keyword searching, where they have been seen as canceling one another out, i.e. the greater the precision, the lower the recall, or the greater the recall, the lower the precision. A good discussion of this is at: http://www.tbray.org/ongoing/When/200x/2003/06/22/PandR. Essentially, a keyword search in a full-text database for "baths of Titus" could bring up information on the Baths of Titus here in Rome (which is what I would want) but it could--who knows?--bring up items about growing fish in India or pornographic sites. Google has gotten around this with their Page Rank system. The traditional method for evaluating human indexing uses different measures: specificity and exhaustivity. If you search human-created indexing terms you would never get the completely out-of-bound results mentioned above (unless the human could not read the text at all), but there can be other problems. The example I use (from my own practice when I was still learning) is a book I had to catalog about the legal rights and responsibilities of pregnant women and new mothers in the Soviet Union. I found copy from a well-known law library that will remain nameless, and found the single subject: "Women--Soviet Union." While it is not as out of bounds as what you might get in a full-text keyword search, it is still wrong from a human indexing point of view. So, we can make it a bit more specific, but even if we put in Pregnant women--Soviet Union, that would not have been specific enough because we have to add the legal aspects. But if we left it there, it still would not be sufficiently exhaustive because we need something for new mothers. I have seen several people mix up the evaluation of human and computer indexing, and the page I gave above appears to do just that. Or perhaps the official definitions have changed and I'm just behind the times, I don't know. But I still don't believe that instituting FRBR or RDA will have any effect on either precision/recall or specificity/exhaustivity. James Weinheimer
