Thanks for that gracious introduction, I very much appreciate it! I’m Dorothea Salo, and I am not nor have I ever been a cataloger, though I may be one of the last generation of library-school students required to take a full course in cataloging, I don’t know. I just know we at the iSchool are doing a major curriculum revision, and current odds are that we’re keeping the core organization-of-information course, just changing it to de-emphasize MARC cataloging in order to include more material relevant to non-MARC environments. Interesting times, not that they’re ever not.
Curriculum revisions are never fun. They eat everybody’s time for months on end, they always cause bureaucratic hassles out the wazoo, and for a while you have to deal with two different incompatible curricula and remember which rules apply to which student you’re advising… it’s a mess. Why do we do this to ourselves? Why do we bother, if it’s such an awful hassle?
My answer is quia faciendum est, “because it must be done.” (Just in Latin it sounds cooler.) We at the iSchool can’t just sit back and do what we’ve always done because we’ve always done it that way, not when the world our graduates will need to fit into is way different from what it was when we built the old curriculum. It doesn’t mean we did a bad job on the old curriculum; I don’t think we did! It’s just that the world has changed out from under it. So we have to change the curriculum. We don’t have to enjoy it. We just have to do it, because it must be done.
Not coincidentally, that’s pretty much how I feel about the move away from MARC. It feels to me like a lot of the library profession has spent half a decade now, at least, on the question cur faciendum est? or “why must it be done?” And even though this question sounds really cool in Latin, I have completely run out of patience with it. Don’t even tell me nobody’s asking this still. I straight-up heard it just last May at a conference; it’s totally still out there floating in the water.
You like water? Here’s some water in a nice harbor, really pretty, love the lighthouse, but the thing that isn’t in this picture of a pretty harbor is a ship. Because the “why can’t we still use MARC?” ship has sailed, people! It has sailed. I’m not even having the why-do-we-have-to-change discussion today, I honestly don’t see the point, that ship has sailed.
I’m next year’s program planner for the IT Division of Special Libraries Association, and I was talking to SLA’s technical-services planner Betty Landesmann—some of you probably know her—I was talking to her about a linked-data session, and she rolled her eyes at me and said “can we not do another intro to linked data and why it’s better than MARC please? I’ve seen a ton of those and they don’t help.” Okay, then. If I’ve got catalogers yelling at me not to do this, I won’t do it! Because really, the answer is exactly the same as it is for our curriculum revisions at the iSchool. Quia faciendum est. Because it must be done.
I do want to mention, though, because I come at technical services from an XML-and-metadata background rather than a MARC background, that it isn’t just MARC cataloging the bell is tolling for here. There’s zero chance XML-based metadata practice will stay the way it is today; I already see it changing. I’m not even sure XML will stay alive as a pure metadata format, as opposed to uses like TEI for the digital humanities and EAD for archives, where you’re dealing with narrative-type documents intended mostly for human beings. I’m okay with XML’s decline as a metadata serialization. I never liked my nice elegant document standard getting worked over by the data engineers anyway—do not even talk to me about XML Schema; it’s just horrific. Maybe now I can have XML back. For documents. As it should be.
No, I’m much more interested in this question: quid faciendum est; what actually is it that must be done? What do we have to do to our catalog data and metadata so it works in this world where so much has changed about how we find information? I like this question because it’s pragmatic, I like it because it’s intriguingly complicated, I like it because it’s nerdy in all the best ways, I like it because I am an inveterate fiddler-with-things and there’s just great huge masses of MARC and XML right there to be fiddled with… and it’s another of those questions we have to work on or we just stay stuck, right?
I don’t think it’s enough to just say “well, we have to migrate our data from MARC and MODS and METS and the various Cores—Dublin Core, Darwin Core, VRA Core, PBCore and so on—we have to migrate all that to linked data.” That’s skipping all the steps! That’s like saying “we have to pick up some rocks and turn them into a giant mosaic.” Whoa, wait, not enough information! What’s our mosaic design? Where will the mosaic be built? Where do we find the right-colored rocks, and how many rocks of each color do we need, and how do we cut them down if they’re too big or not the right shape? How do we glue the rocks down? What if somebody makes a mistake? What if there’s an earthquake?
Process. That’s what we need here. Some process, right?
The other reason I don’t think it’s enough to just say “well, we have to migrate our data from what we have to linked data” is that it assumes without proof that linked data is the ultimate destination for it. It’s the horse to bet on, I’m not saying it isn’t, but I just think linked data gets used as a stalking horse sometimes, a scapegoat. It’s linked data’s fault we can’t use MARC, it’s linked data’s fault all these changes are happening, if it weren’t for linked data supposedly being the new hotness we could stay the way we are and everything would be fine.
I don’t actually think that’s true. If linked data didn’t exist—and let me tell you, I have lots of days I’d love to wipe RDF off the face of the earth—if there were no linked data, we’d still have to make changes in how we collect and organize our catalog data and our metadata. We have to make those changes for the same reason we’re changing the LIS curriculum at the iSchool: the world has just plain changed out from under the old ways. That didn’t happen when the Library of Congress or the British Library announced their linked-data plans. It happened long before that. It happened when paper cards gave way to the Web as the main way patrons interact with library catalogs. It’s not that we didn’t notice; of course we did. It’s just taken us a while to figure out what we need to do about it. I don’t know how to say “oops” in Latin, but oops, I do think we maybe waited longer than we should have. Water under the bridge now.
That leads me right back to quid faciendum est? What do we have to do, now that the work we do has to play nicely with computers? And not just computers—MARC was designed for computers—but networked computers, computers that can talk to one another. The network really does change the game.
When I teach our core organization-of-information course, I start my students with this thought: ordinatra stulta sunt. Computers are not all that bright, I tell my students; you’re way smarter than a computer. I do this for a lot of reasons. One reason is knocking computers off pedestals—not actually literally knocking computers off pedestals, though hey, that would be kind of awesome—but you know what I mean, right? A lot of my students come into the iSchool thinking that computers are like unto small gods, magical and capricious and liable to mess you up, impossible to understand much less work with. I’m saying, I have to get them to not think that, because the more they understand about how computers do work, the better off they are, and the better off we all are.
But the main reason that the notion that computers aren’t real bright is relevant to this talk today is that from our point of view as literate human beings, computers are not-too-bright in some very specific and fairly easy-to-understand ways, and those ways tell us pretty clearly what our catalog data and metadata have to look like if we want computers to work effectively with it. Really this is no different from how the shape and size of catalog cards and the standard size of typewritten lettering shaped how the MARC record had to look. The technology you have available—and the card catalog is totally a technology, don’t let anybody tell you it isn’t—the technologies available to you shape how it makes the most sense to do things, because different technologies are good and bad at different things and need different things to function best. That’s basic design theory; read Donald Norman’s Design of Everyday Things and its sequels if you haven’t already, it’s brilliant stuff.
So the first thing to remember about computers is that text, the ordinary stuff we write for other people to read—the text we literate human beings read and comprehend so fast and easily that we hardly have to think about it—it’s all Greek to a computer. (You knew that cliché was showing up at some point, right?) Computers can’t read; they are functionally illiterate. If anybody in this room has a kindergarten-aged child at home, that kindergartener most likely reads and comprehends text immensely better than a computer can.
So, in my head, one step toward coping with illiterate computers is dealing with our addiction to textual notes. By way of example, I took some MARC 504s straight from the Library of Congress’s MARC documentation—thanks for that, by the way, if anybody here is responsible for it, it’s super-helpful in my classroom:
- 504 ##$aBibliography: p. 238-239.
- 504 ##$aSources: p. 125-152.
- 504 ##$aIncludes bibliographical references.
- 504 ##$aIncludes bibliographies and index.
- 504 ##$aChronological list of the author’s works: p. 469-475.
- 504 ##$a”Selected bibliography”: v. 1, p. 351-358, v. 2, p. 234-236.
- 504 ##$a”Spisok izdannikh knig”: p. 277.
So, for example, if a patron question we would like our catalogs to answer is “hey, I’m new to this topic, can I get a recent book with a good bibliography please?” these notes being free text means our catalogs can’t answer that question, because to get to an answer means filtering a list of books by whether they have a bibliography or not, and to do that with MARC notes, a computer has to understand that “bibliography” and “bibliographies” and “bibliographical” and “literature cited” and “sources” and maybe whatever that romanized Russian means—I didn’t look it up—the computer has to figure out that they all mean yes, there’s some kind of bibliography! And no, a computer can’t just look for the existence of a 504 and assume there’s a bibliography, because some 504s don’t say anything about bibliographies, just indexes.
Look. The computer is not bright enough to figure this out. It can’t read, much less read all the languages we transcribe stuff in, much less comprehend what it reads. That makes a lot of the stuff in our MARC records a lot less useful to patrons than it could be.
Computers do understand yes or no; that they’re real good at:
- ☑ Bibliography
- ☑ Index
Checkboxes are candy to computers. So for any conceivable criterion we want our patrons to be able to filter their catalog results on, we pretty much have to quit recording it in text and make it a checkbox—or radio buttons if there’s more than two options; that works too. If you get the sense from this that I like MARC fixed fields, you’re right! (Though honestly, that festschrift thing? That is just weird.) If a lot more of MARC had been expressed in fixed fields instead of free text, we’d be a lot better off right now.
Now, let’s say for a moment that the titanic arguments we in the profession will have to have about when a book can be said to have a bibliography, and what counts as an index… let’s say those are over, and we’ve drawn the best line we can. We still have to deal with this giant horrible mass of free-text notes in our existing catalogs that computers aren’t bright enough to understand. Faciendum est, people, it’s got to be done. I don’t want to get down in the weeds on this—actually I would totally love to because this is exactly the kind of problem I salivate over, but I’m a giant nerd and not everyone is like me in that. So I just want to say, this class of problem can be solved for the great mass of our records without hand-editing, and of course it must be, because there ain’t enough staff time in the universe to check all those checkboxes. And it’s a thing that must be done for every characteristic in our notes that we want users to be able to filter or search on.
Another serious and ugly free-text problem we have in our records has to do with places where our content standards don’t force us to be consistent about how we record certain information. Quoth the mighty AACR2r, “You know, whatevs. We’re cool. Just type something.” On catalog cards this inconsistency didn’t matter, because the information was only ever going to be skimmed by a human being, who doesn’t need a whole lot of consistency. We’re literate humans; we’re smart about figuring this stuff out. Computers are astoundingly literal-minded, however. You can take some text, add a space to it, and to the computer it’s suddenly something totally different. Just one space, that to a human isn’t even visible!
Once again, this isn’t just a MARC thing. Consistency in Dublin Core metadata? Don’t make me laugh. Actually, I’m more likely to cry. It’s bad out there, people. I mean, I once got an entire published article out of one poor soul in the institutional repository I was running at the time—poor guy had stuff under eight slightly-different spellings of his name. Just terrible, and yes, I fixed it as soon I had the screenshot I needed for the article—of course I did, I wouldn’t just leave it there in that state—but it goes to show.
I want to call out two Dublin Core things specifically, noting that you’ll find these problems many more places than just Dublin Core. Dates, oh my goodness, dates. Dates are really important to information-seekers, so it’s really important that we record them consistently such that a computer can reasonably intelligently filter based on them. We are not there yet; we’re not even close to there. Dublin Core, MARC, it doesn’t matter, people who try to make computers work are tearing out their hair about the ways we do dates, and we have to fix it. Faciendum est, we have to fix it.
Rights statements also come from the digital-collections side more than the catalog side. It’s important because we have to be clear about what our users are allowed to do with our digital collections, and to do that, our search engines have to be able to tell which users can do what with which items, and that’s all free-text now and it’s a total mess. Europeana and the Digital Public Library of America are working on it, thankfully, though I don’t envy them that job one bit.
An example of the ugliness of free text that I use in class a lot is from library software developer Bill Dueber, who took a close look at what was after the ISBN in the 020 field in the catalog he was working with. It’s horrific. Just the top twenty responses by number of occurrences, you can see the inconsistency:
- 402537 pbk
- 387406 alk. paper
- 99260 v # (e.g., “v. 1”, “v. 22”, etc.)
- 82918 cloth
- 51125 hbk
- 42036 electronic bk
- 41360 acid-free paper
- 38792 hardcover
- 28913 set
- 20358 hardback
- 19160 ebook
- 16264 paper
- 15269 u.s
- 12770 hd.bd
- 11793 print
- 10625 lib. bdg
- 10520 hc
- 8772 est
- 7767 pb
- 7639 hard
The more you drill down, the worse it gets. Dueber looked for everything that might reasonably represent a hardcover book:
hard | hard adhesive | hard back | hard bd | hard book | hard bound | hard bound book | hard boundhard case | hard casehard copy | hard copy | hard copy set | hard cov | hard cover | hard covers | hard sewn | hard signed | hard-backhard-backcased | hard-bound | hard-cover | hard-cover acid-free | hardb | hard\cover | hardbach | hardback | hardback book | hardback cover | hardbackcased | hardbd | hardbk | hardbond | hardbook | hardboubd | hardbound | hardboundhardboundtion | hardc | hardcase | hardcopy | hardcopy publication | hardcov | hardcov er | hardcovcer | hardcove | hardcover | hardcover-alk. paper | hardcovercloth | hardcoverflexibound | hardcoverhardcoverwith cd | hardcoverr | hardcovers | hardcoversame | hardcoversame as above | hardcoverset | hardcovertion | hardcver | hardcvoer | hardcvr | harddback | harde | hardocover | hardover | hardpack | hardpaper | hardvocer | hardware | hd | hd bd | hd. bd | hd. bd. in slip case | hd. bd.in sl.cs | hd. bk | hd. cover | hd.bd | hd.bd. in box | hdb | hdbd | hdbk | hdbkb | hdbkhdbk | hdbnd | hdc | hdcvr | hdk | hdp | hdpk | hradback | hradcover | hrd | hrdbk | hrdcver | hrdcvr
So our catalogs can’t answer the very simple question “yo, this book, print or electronic or both?” At least not based on the 020—and yes, I know RDA fixes this and I’m pleased about that. Bottom line, though, a lot of catalog data is hopelessly internally inconsistent. Sometimes that’s material for patrons and sometimes it isn’t, but when it is, this is my call to control all the things. All of them. Anything useful in a record that isn’t actually transcribed off the item needs a controlled vocabulary, or other appropriately-standardized expression if it’s something like a date. I cannot with this nonsense, and neither can computers. “Whatevs, just type something” is not okay in 2015. Transcribe it or control it; there is no third option. Faciendum est. Oh, and since I’ve said the word “transcribe,” let me just say, intentionally transcribing typos and other errors in information that’s material to a patron’s searching and browsing is completely mindboggling to me. We have to fix that stuff, and stop propagating mistakes. Consider it a service to publishers, as well as our poor patrons.
I picked on the 020 field for another reason too, having less to do with cataloging practices and more to do with ISBNs. Now, I know I can’t fool y’all the way I fool my org-of-info students with the question “is the ISBN a good book identifier?” We know it’s not. We know lots of books don’t even have ISBNs, and sometimes ISBNs get repeated for different books. It’s not totally clear what a “book” even is in ISBNland; it’s kind of an edition but not really, it’s kind of a format question but not really, and it’s all very confusing. Perhaps predictably, it’s confusing to computers too. Computers need to be really super unambiguous when talking about what kind of thing something is—if you and the computer have a different definition of what a “book” is, the computer is going to do random unexpected and unpredictable things from your point of view. The computer is happy to use whatever definition or definitions we’re happy with, the computer doesn’t care… but in spite of FRBR and sometimes (it must be said) because of it, we don’t really have clear definitions here that don’t lead us into logical contradictions or bad edge cases.
So that’s one thing. We have to figure out what exactly we’re talking about when we say things like “book” and “ebook” and “hardback” and so on, so we can explain the distinctions clearly to the computer… and if this reminds you of Suzanne Briet trying to explain when an antelope is a document and when it isn’t, I am right there with you! It’s totally going to be weird and sometimes theoretical like that.
Once we know what kinds of things we’re talking about, we have to be able to point clearly and unambiguously at every single example of these things that we—we collectively—have, so that it’s easier to pool the information about these things that we collectively have. The network can work for us if we let it—all of us know more than any of us about what we all have—but to let the network work, we have to have a common way to point at things. And for a computer, that means an identifier that (unlike an authority string) never, ever changes. And for a networked computer, that means an identifier that’s unique not just in your organization—so no, your call number or barcode number won’t work—but unique worldwide, so it absolutely cannot be language-dependent.
We’ve tried collating our records by fuzzy-matching and deduplicating; that’s how metasearch worked. But we pretty much all know that metasearch never worked real great. Computers aren’t bright enough to fuzzy-match well, and catalog data and metadata are sparse enough that they’re not good candidates for that approach to begin with. We’ll still have to use it to assign identifiers to start with, because we don’t have anything better, but it’ll be a long haul for some stuff and some catalogs. So that means we need unique identifiers for our stuff that are way more reliable than ISBNs. If you know the linked data world at all, you know that the scheme that’s been settled on for these is URIs, which mostly look like URLs. The reasoning there is that we already know how to make URLs globally unique, because they already have to be or the web doesn’t work. That’s all it is, just keeping the computers from getting confused.
Another problem we have to solve if we’re going to take advantage of the network is taking identifiers and quasi-identifiers that we’ve been using for things, like ISBNs and authority strings, and matching them up with URIs/URLs that have been established for those things. Again, I’m not going down in the weeds here, but I do want you to know (if you don’t already) that the nerd-word for this process is “reconciliation” and it can be partially automated if you know your source data well, as catalogers generally do.
Once you have a URI for something, you can go out to the network and ask a whole bunch of trustworthy sources “hey, what do you know about this thing?” and get back useful answers. To me, that’s how what we now think of as copy cataloging is going to work. Ask tiny questions of various reliable sources, get tiny answers, build them up into search indexes and facets and browsing tools and all the other UI chrome we’re already familiar with. It won’t have to be done by hand, if you tell a computer “every time I feed you a URI for a book we’ve bought, ask this question of that source in this way and store the answer there” it will happily do that, reliably and consistently, every single time. I strongly believe this will be a much better solution to what I think of in capital letters as the Problem Of Vendor Records. You’re familiar with this problem, I don’t have to elaborate, right? What I’m saying is, the Problem of Vendor Records is nine times out of ten a problem of vendors struggling not only with MARC and AACR2, but with MARC practices that are incredibly inconsistent across libraries. It’ll be a lot easier for us and for vendors if instead our computers ask their computers a lot of tiny questions with tiny answers.
That leads to one last thing about the 020, okay? ISBNs aren’t unique, the field includes inconsistent format information, yeah, yeah, we got that. Here’s my question: what the everliving heck is format information doing in an ISBN field to begin with? Much less information about volumes of a multi-volume series? And if it does have to be there—and I know, I know, I do understand why it ended up there—why isn’t it at least in a separate subfield? What’s this nonsense with parentheses?
I’m showing you a mosaic detail here for a reason. You can see all the teensy-tiny individual rock bits here, and you can see how carefully they’re placed, and that none of them actually overlap anywhere. That’s what our catalog data and metadata should look like. No overlaps, nothing jammed together, everything in tiny tiny bits and each tiny bit in its own singular place, very carefully set apart from all the other tiny bits. That’s called “granularity” or “atomicity,” and computers love it. Computers are really really good at building up whole mosaics from tiny granular pieces! What they’re critically bad at is taking a whole and breaking it into parts. We have to do that for them. And as we saw with the 020 field, we often don’t, or when we do, we do it in ways that the computer finds confusing and inconsistent. Here’s a relevant factoid to take home with you: Computers cannot reliably and consistently split human names and titles in human languages into their component parts. Naming is just too inconsistent across human cultures and languages for that to work. With pre-split names, though, it’s relatively easy to write rules to put the names back together intelligibly.
So I grabbed a couple of examples from IFLA’s ISBD supplement—and again, thank you for this, if anybody’s here from IFLA, this is a classroom life-saver—just to show the difficulties:
Object (cartographic ; visual) : unmediated
Scanglobe diameter 30 cm / by Replogle. — Scale
1:42,000,000. — [Chicago, Illinois] : Replogle
1 globe : col., plastic ; 30 cm (diam.)
Relief shown by shading and spot heights. Depths shown by shading and soundings. — “Scanglobe” is a trademark. — Mounted on spindle crowned by a plastic clockface, in a plastic meridian half circle, on plastic base. — Globe lights up from inside by means of electrical cord with switch button and interior bulb
Try to think like a computer for a second. There’s a whole lot of punctuation all up in here, and it’s not at all obvious (even to me, and I’m a human being) what’s being set off by it or what it means, or even if it means anything at all! I mean, look at the physical description of the globe there. Tell me what a period, a dot, means there. As a human being, I can figure out the dots there are calling out abbreviations. Now, can I just tell the computer to assume a dot always means an abbreviation? Of course I can’t! A dot doesn’t mean that in the other areas! And can anybody tell me why in the last area of the globe description, everything except the final sentence has a period at the end? It’s enough to make a computer programmer cry into her beer.
I lopped the area labels off for legibility, but I just want to point out, we have two competing sets of delimiters happening here—the areas, that are set off with whitespace, and what’s in each individual area, which is funky punctuation city. And when you add that to MARC, we’ve got a whole ’nother set of delimiters in the form of fields, subfields, and indicators. I respect my colleagues who teach cataloging! I could never do it! Because I cut my teeth on XML, where delimiters are totally cut-and-dried and straightforward, such that I find this mishmash completely bewildering!
A lot of my students find learning HTML and XML frustrating, because they’ve never had to be one hundred percent consistent about delimiters before. So they make tiny little mistakes like leaving off an angle bracket, and they haven’t learned to scan for those mistakes yet, and they don’t understand what the validator that does notice problems like that is trying to tell them, and it’s really frustrating for them. What I tell them is, suck it up and deal. Okay, no, I’m not actually that evil about it. I’m pretty careful to point out the kinds of errors that beginners usually make, and I tell them that everybody makes those errors, even really skilled and experienced people, and it’s okay, the whole point of validators is to help us get it right. But fundamentally, they have to learn to deal, and so do we. They don’t like that necessarily, and we don’t have to like it either, but too bad, faciendum est, it must be done. Don’t confuse the computer, folks! Reliable and consistent delimiter use is how we avoid confusing the computer. Delimit in just one way, delimit clearly, delimit unambiguously.
Even our XML-based metadata standards aren’t necessarily doing that, much less our MARC-based catalogs! MODS has composite fields and weird ISBD punctuation. I have to share this thing that broke my heart—I was digging into the history of MARC and ISBD internationally, and it turns out that the Germans were totally bent on killing ISBD punctuation out of MARC and relying on MARC delimiters only, which from the point of view of 2015 totally would have turned out to be the right decision, but English- speaking MARC went the ISBD direction instead, and just argh, 20/20 hindsight. One more mess we now have to clean up.
As a recap, here’s what we have to do: We have to get a handle on our free-text issues. When we’re saying the same thing, we need to say it the same way every time. We need to atomize our data, make it as granular as it can be. The delimiter thing, sometimes we have too many and sometimes we don’t have enough, and we need a happy medium. And when we can identify something as well as labeling it, we should, because identifiers make computers happy and useful.
Just to reiterate, we don’t have to do these things Because Linked Data. We have to do these things Because Databases, and Because Search Engines, and Because Faceted Browsing, and Because Internet, and Because Web. Basically, Because Not-Too-Bright Computers. Now, no secret, cards on the table: if we do these things, we’ll be one heck of a lot closer to having linked data. So it’s not like I’m ignoring that, I’m just saying linked data is not all this is about.
Well, that’s great. Plenty of work for catalogers, right? How do we actually do it, exactly? Quomodo faciendum est?
“By hand, one record at a time,” is not the answer here. It can’t be. I mean, yes, there will still be weird outliers that we end up fixing by hand—there always are, I could tell you stories—but we need to throw computers at fixing the easier problems and limit our handwork to those weird outliers. We have too much data to do it any other way. I’m not sure that doing it organization by organization is the way either. In my head, that means a lot of the same problems are getting solved redundantly in parallel. That costs too much and takes too long.
But the truth is, we collectively don’t yet have the know-how we need to collaborate on this. Most computer programmers don’t have the MARC knowledge, and many catalogers don’t know how to work with records en masse. So this problem? Solvitur discendo; it is solved by learning. Catalogers, developers, we all have some learning to do. I would say that, right? I’m a teacher. But I’m a teacher because I believe this.
Any managers in the room, supervisors? Give your people time and space to learn; do not make me yell at you about this! Your cataloging backlog or the new tech thing your developers are working on is far less important than your strategic preparation for what’s barrelling down the pipeline at you, okay? By all means hold your people accountable for actually learning as opposed to complaining about learning—again, I could tell you stories, but you can probably tell me stories too—but let them learn. Help them learn. Learn yourself, it won’t kill you and might make you stronger.
Here are some tools I think are well worth adding to your toolkit if they’re not there already, because they’re designed to fix stuff in lots of records at once rather than one-record-at-a-time. I’ve ordered them by the order in which I would recommend that a cataloger learn them. The last half you may not even need; I mention them because there are situations where they’re genuinely going to be useful.
- OpenRefine (try LODRefine)
- Regular expressions (try regexone.com)
- SQL (if you have to extract data from a relational database)
- XSLT (if you have a lot of XML around)
- Catmandu/Fix (don’t start here, Here There Be Yaks)
How should you learn, and what should you do with what you learn? In my head, those questions are intimately entwined. In some circles I’m known for the phrase “beating things with rocks until they work.” If you use mosaic rocks this probably won’t work, but you get the idea. I think plain old mucking around and breaking things and fixing them is the best way to learn new tools, myself—it’s certainly how I do it, and it’s the method I have my students use. Pick something to do with the tool, then do it, and if a few rocks get beaten on along the way, it’s all good.
What should you do exactly? Well, try cleaning up your catalog data! Go fix your 020s with MARCEdit, see if you can at least make those format notes more consistent. Export your data into something OpenRefine can read—it’s doable, I’ve talked to catalogers who did it—and see if you can cluster your 504s such that you can figure out in the majority of cases if there’s a bibliography, if there’s an index. Try fixing your dates, there are all kinds of fun and interesting problems you’ll run into and have to come up with some way to solve. You know your catalogs way better than I do; you know where the worst problems are. You also know where the most important problems to fix are, which I think is totally crucial knowledge here. So learn the tools by fixing the problems you already know are there.
We’ve got a lot of work to do, and it’s got to be done, so let’s jump in and do it. Thanks very much, and I’m happy to take questions.