Soylent Semantic Web Is People!

So hello, guten morgen. My name is Dorothea Salo, and I teach many nerdy things, linked data among them, at the School of Library and Information Studies at the University of Wisconsin at Madison. I first want to say vielen dank—thank you very much—for inviting me here, and I hope I can kick off this conference in a fun and useful way.

It’s the year 2022… People are still the same.
They’ll do anything to get what they need.
And they need SOYLENT GREEN.

Soylent Green movie poster

In 1973 Charlton Heston starred in a science-fiction movie called Soylent Green. And it’s a terrible movie, talky and preachy and weirdly acted and often just ridiculous, so I don’t feel too bad about spoiling the big plot twist. In the movie, the environment has degraded so badly that food can’t be grown, so what everybody eats is artificial foods called Soylent Whatever—Soylent Red, Soylent Yellow, and the brand-new Soylent Green. What they don’t know, until Charlton Heston yells it at the end of the movie, is that Soylent Green Is People! More specifically, Soylent Green is what happens when you make people into food. Ew. But the total nastiness of cannibalism aside, what’s interesting about this movie is that you have this whole society that has absolutely no idea that it’s completely dependent on people for its survival!

It’s the year 2013… Data are still the same.
We’ll do anything to make sense of them.
And for that we need PEOPLE.

Now, we’re not cannibals here in Hamburg; we don’t actually eat people. The parallel I want to draw here is that the original Semantic Web vision curiously lacked people, except maybe as the end-user beneficiaries of linked data. I mean, you can go back and look at what Berners-Lee and his cronies wrote, and you have all these people booking travel and getting health care or whatever because of all the nice clean shiny RDF data whizzing around in nice clean shiny server rooms, sure. But the data whizzes around all by itself. Doesn’t need people. There are no people. Just data.

I just think this is a counterproductive, even dangerous, way to frame the Semantic Web—and it’s still much too common. So I assert that the Soylent Semantic Web Is People! Because I want a human semantic web. A humane semantic web. Technology without people is just dead metal and silicon. Data without people is just noise.

It’s the year 2013… Data are still the same.
We’ll do anything to make sense of them.
And for that we need LIBRARIANS.

And more, since we’re here at Semantic Web in Libraries, I will assert that Soylent Semantic Web Is Librarians! We are the Semantic Web, and the Semantic Web is us! I know that isn’t completely news—we invented SKOS, we invented Dublin Core, we have Karen Coyle and Diane Hillmann and Ed Summers, just for starters—but if you had to ask me why this specific conference is important? That’s what I’d say. The Soylent Semantic Web Is Librarians.

What I want to do today is explain my thoughts about why the Semantic Web is not soylent, not made of librarians, not made of people. I want to explain why it should be soylent. And I want to challenge you in specific ways to make it soylent. My ultimate goal, which I imagine you share, is strengthening library adoption of linked data.

Let’s decide, in approved RDF-triple style, just what properties we can assert about librarians and linked data. And the usual properties I would expect people at this conference to suggest would be the technical ones. Librarians model linked data. Librarians crosswalk to linked data. Maybe as simple as librarians make linked data. Librarians host linked data. Librarians archive linked data. Librarians build systems for, and around, linked data.

But none of those properties really belong to the Soylent Semantic Web, the Semantic Web made of people. These properties are about the data, not the people.

Here are some things librarians do, as people, in the Soylent Semantic Web. We investigate linked data. We discuss linked data, sometimes not as knowledgeably as linked-data advocates might like. We learn about linked data. We teach about linked data. We advocate for linked data. Or don’t. And now we get to the crucial point: we adopt linked data.

Or we don’t. We mostly don’t adopt linked data, in fact, and we don’t because the Semantic Web community, librarians included, hasn’t acknowledged that it needs to be soylent. We forget that the Semantic Web is made of people, lots of different kinds of people, some of them people who are not like us and do not do the same work we do and do not have the same understandings we have. We forget that we need our own librarian colleagues to help us make the Semantic Web, and put library data into it—and when we forget our librarian colleagues, our librarian colleagues forget us, and forget linked data. And that’s not good.

As I talk to librarians about linked data, what I hear back is that they feel ground up into hamburger—sorry, sorry, I had to—by the whole thing, because the way it’s usually explained to them, it’s so abstract and so divorced from the actual library work they know. The linked data movement can show them graphs, but it can’t show them interfaces for doing their work. It can tell them about triples, but it’s not telling them how the catalog will work if their Internet connection fails. It can explain ontologies, but not how they’ll navigate them.

After one explanatory talk I gave, I had one cataloger tell me with immense frustration, “I just don’t see how this will work!” And I didn’t have a good answer for her, because I don’t see that either.

This has happened before.

Now, switching away from Soylent Green briefly to—anybody recognize this? I took it from the remade Battlestar Galactica television series, which uses it as a catchphrase. This is not the first time an upstart technology has tried to upend an entire established infrastructure, along with the people using it.

At the turn of the century, I was working in publishing: specifically, electronic publishing, and even more specifically, ebooks and electronic journals. While some of the big journal publishers climbed onto the XML bandwagon, many other journals didn’t, and the trade publishing industry just never did. I remember sitting in an ebook conference next to a high-level editor from a Big New York Publisher, and we were listening to a fairly basic, fairly standard introduction to XML, and I heard her sigh “This is just not my world any more.” She felt alienated. She felt alien. Is there anybody in this room who hasn’t heard a colleague express that alienation?

Even worse, XML didn’t make publishers’ lives easier—it made them harder! Editing, typesetting, indexing, all these workflows got hugely more complicated for what looked at the time like super-dubious returns. And the XML community took no notice whatever of their difficulties, the difficulties actual people were having doing actual publishing work with XML. Why? Because the XML community was having way too much fun loudly proclaiming XML’s superiority over everything ever, and going off into corners to have arcane technical arguments about XML namespaces. Not very soylent! Not humane! Not made of people!

Now, publishers did still make some XML, I grant you. I saw a lot of it. Forgive my language, but trade publisher XML was crap. It was garbage. You wouldn’t feed it to your pet Cylon, it was so bad! Which goes to show that technology that doesn’t fit into real people’s environments won’t be used properly, if it’s used at all.

How many of you knew a slide about institutional repositories was coming? Go ahead, raise your hands. If you know me, you know that I am just so sad and angry about institutional repositories. In Europe, I know, it hasn’t been quite so bad, but in the States, it’s been wretched. But it was the same thing again. There was this technology that was going to make everything better, only the people making the technology forgot all about the people who were supposedly going to use it. So we got these lousy unusable unfixable systems that did lousy useless things, and no big surprise, nobody willingly put anything in them! Because they weren’t soylent! They weren’t made of people!

Incidentally, what happened to the people running institutional repositories? People like me? Well, we got blamed. And I, for one, got out. I will never work on an institutional repository again. This is a thing that happens when systems don’t treat worker-people well. Worker-people abandon those systems, even people who truly believed in them and had high hopes for them. So when we linked-data folks lose catalogers, I think it’s a serious problem.

This will happen again.

Battlestar Galactica

We have plenty of history of technologies not succeeding because they aren’t people-conscious enough. This will happen again to linked data, if we’re not careful, and if the Semantic Web doesn’t remember that it’s soylent—made of people. I don’t want that. You don’t want that. But that’s what’s going to happen if we can’t bring more people to linked data.

It’s the year 2013… RDF is still the same.
Why do people who should know better still believe
RDF is based on XML?

Just as an example, I was at ASIST a couple of weeks ago, the big annual conference for the Association for Information Science and Technology. And I went to a session on linked data—and I won’t be any more clear than that, because I’m not here to embarrass any specific person—and I saw this on a slide: “RDF is built from XML.” This kind of thing makes me think that eating people alive might actually be an interesting lifestyle choice! Maybe you too? Because my gosh, it’s 2013, RDF never was built from XML, so why on earth do people who really should know better still believe this strongly enough to put it on a presentation slide?!

Clearly education, even really basic education, is a problem here. It’s a people problem, not a data problem, and as an educator, it’s my problem, right? I think of education as my major role in furthering the adoption of linked data in libraries: educating future librarians and archivists and other information professionals, and educating current ones, which I also do. I have to tell you, though, that current linked data infrastructure is not making this easy for me.

Give me 45 minutes, and I can drag a roomful of complete HTML novices through making an extremely basic web page. I know this because I’ve done it! Give me another 45 minutes, and I can drag those same people through the basics of CSS. Again, I know this because I’ve done it. No, they won’t be web designers after that, but they can go and practice usefully on their own and get better, and there’s a ton of resources on the web to help them. XML is a bit harder to explain and work with, but if my roomful of people is actually a roomful of librarians or library-school students, I can drag them through being able to make a basic MODS record in two and a half hours or so. I know this. I’ve done it.

Here’s the thing. I don’t know how much time it takes to drag a roomful of novices through minimal RDF competence. I’m not even sure what minimal RDF competence looks like! So essentially it might as well be infinite time. I’ve tried, I really have. I just don’t think I’ve succeeded. What are the problems I’m running into?

Part of my problem is that the training materials I have to work with force my librarian learners into stunts like trying to catch a ball while jumping off a diving board. Really, a lot of the stuff that’s out there, even I bounce right off of—and I supposedly know RDF well enough to keynote a semantic-web conference!

Take this linked-data introduction from Cambridge Semantics—and in fairness to them, they didn’t make this for librarians, but it’s still one of the best things out there. But look at it: just the first sentence and we’ve already brought in HTTP and TCP/IP without defining them, much less explaining why they’re important in this context. My learners? My librarians and library-school students? They don’t know about the alphabet-soup plumbing of the Internet. They might have heard HTTP and TCP/IP mentioned (quite likely by me, in another class), but that doesn’t mean they know. They’re just going to bounce right off this, or get distracted by something that’s actually a pretty minor and useless detail.

It gets worse. What’s the metaphor this introduction picked out, to explain linked data? The relational database, speaking of things a lot of my learners don’t know about! So this extremely well-intentioned and well-written tutorial is useless to me. It won’t help the people I have to teach, so it’s not soylent.

The answer to this dilemma is not to call my learners stupid. I warn you, I am not even going to listen to that, so don’t anybody try it. I’m also not going to listen to any suggestion that librarians can’t learn about linked data until they learn TCP/IP and HTTP and relational databases and XML and at least three programming languages. That’s ridiculous. I’ve been teaching tech to future librarians since 2007, and trust me, with most things you can meet them where they are—which can, yes, be a really low skill level—and still teach them a lot.

How does that work? The answer—the soylent answer, the answer that acknowledges my learners’ humanity and their love for what they do—the answer is respect, primarily respect for librarians’ existing knowledge base. This is the principle I try to build my lessons on: draw from what my learners already know. I try to teach linked data based on my learners’ interest in it. No surprise, for most of them, their interest has a lot to do with linked data replacing MARC. The rest of them are digital librarians and archivists, or aspiring digital librarians at any rate, and for them I keep library metadata practices in mind.

So, for the sake of time, let’s just stick to MARC. What happens when I try to translate MARC skills and practices into a linked-data context? What happens is the same thing that happened with publishers and XML—I crash my little linked-data car right into all the work that libraries now do, all the work that forms the foundations of library data, that is just impossible to even demonstrate with linked data.

I won’t tell you all my tales of woe—I have a lot of them!—but here’s one. I teach this continuing-education course that introduces XML and linked data to working librarians. This fall I wanted to add a couple of weeks on Open Refine to it, because I thought that data cleanup was important to teach. And I thought that reconciling some random spreadsheet metadata with existing linked datastores would be a cool demo, with pretty obvious relevance to real-world librarian work.

So naturally I thought about name authority control, because it’s just so basic to what librarians do, and because it’s something the rest of the linked-data world is learning to do from libraries. Even in the States, where we’re kind of behind Europe in linked-data experimentation, we have these great name authority linked-datastores, VIAF and the Library of Congress, so I thought a little reconciliation would be easy.

I learned very quickly, of course, that I can’t use VIAF from Open Refine, because there’s no SPARQL endpoint for it. I’m on the record here, so I’ll just say—you tell me why not. So I said to myself, okay, that doesn’t work, what about the Library of Congress? Naturally I went right to the source, Ed Summers, because who wouldn’t?

Oops. Ed told me I can’t do authority-control reconciliation that way either; I’d have to download the entire dataset and load it into a triplestore offering SPARQL-endpoint capability. This is where I confess the limits of my own knowledge: I don’t know how to build a web-available triplestore with a SPARQL endpoint off somebody else’s data! And this lesson I was working on was two weeks from going live—I didn’t have time to figure it out!

I asked Twitter if anybody else had maybe done authority control with Open Refine and could show me how. I just needed a simple demo!

I heard nothing.

Let me just say that trying to put together a useful lesson about how to do actual library work with linked data was not a super-humane experience. I felt annoyed. I felt stupid. I felt frustrated. I felt like hey, if the Semantic Web is so soylent, how about I just eat up all you linked data nerds? And I am a vegetarian!

Authority control is basic, basic stuff, folks. Many librarians consider it a touchstone of library practice, something central to our professional identities (so to speak). If I can’t do authority control work with linked data, do not even talk to me about how linked data is more flexible, linked data is wonderful, linked data is superior—linked data is useless. It is useless for librarians in practical terms. That’s not a problem with librarians. That’s a problem with linked data.

The end of the story, just to add insult to injury, is that I discovered that Open Refine’s built-in Freebase reconciliation didn’t work. I was able to fix it, after some searching and fiddling, though, and that leads me to another thing I want to talk about, which is the state of tools available for just messing around with linked data.

I am showing you the instructions for installing the RDF extension for Open Refine—which, by the way, I think this is great and I want more things like it. These are the long instructions, mind you—there’s a shorter set on the main page. There’s a major error in these; you can’t actually get to the workspace directory from the Open Refine start page, because the start page starts on the Create tab, not the Open tab. I flatter myself I’m pretty tech-savvy, but I had to click around and swear a bit before I figured out what these instructions were getting at.

I ended up writing my own installation instructions that seemed to work pretty well. You’re welcome. Please don’t make me do this again. Wrong installation instructions are just not soylent, and this installation method is ridiculous on its face, not soylent at all.

If there are better tools—tools that help me help my learners get actual library work done with linked data—I do not know what they are. I’m not sure they even exist. And that’s a gigantic problem for me as an educator, and ultimately it’s a gigantic problem for you and for linked data. If I fail at my job, you know what happens. It’s what happened with XML and publishing, where XML did not help get publishing work done. It’s what happened with institutional repositories, which basically didn’t help anybody get any work done.

Soylent technologies, technologies that are so respectful of people that people jump for joy about using them, help those people get stuff done. It’s as simple as that. And this needs to be true for people who are not linked data nerds and not programmers.

Look, fundamentally, this is the same reason programmers hate MARC! MARC gets in the way of programmers getting useful work done, right? But if linked data puts every other librarian on earth in the position that library programmers are currently in, that’s not going to help linked-data adoption in libraries.

So to sum up here… because I can’t educate people well, and because the tools are so bad, and because practically nobody can actually get library work done with linked data, linked data is stuck in what I’ve seen called “negative path dependence.” What’s negative path dependence? I quote from a recent report on data sharing: “Because of high switching costs, inferior technologies can become so dominant that even superior technologies cannot surpass them in the marketplace.” Sounds like XML in publishing, right, compared to PDF? Sounds like institutional repositories against journals, right?

I’m afraid it sounds like linked data against MARC, too. Meaning no disrespect at all to the great Henriette Avram, MARC is the inferior technology here—I really believe that! But linked data, despite its superiority, can’t get library work done at this point without ridiculous costs, so it can’t replace MARC.

But it doesn’t have to be this way. This I also believe.

I’ll close with four challenges for the Soylent Semantic Web, the Semantic Web that is made of librarians and other people. I hope—and I believe!—that presenters at this conference will answer these challenges, and I look forward to seeing that… and I also hope that all of you take these challenges home and work on them.

Challenge #1: Work, not ontologies, for linked data

Here is my linked-data heresy. Feel free to turn me into hamburger for it later: I don’t care about your ontology. I don’t care about anybody’s ontology, or data model, or graph, or whatever. I do not care. Why should I? We’ve done library work without ontologies and picture-perfect data models for hundreds of years, somehow or other. Can we just get off ontologies already?

What do I care about? I care about the work I can do with linked data, and the work librarians can do with linked data, and the work my learners can do with linked data. I care about the tools that help them do that work. I care about the work skills I can realistically teach my learners that someone will pay them for—and before you say anything, “knowing an ontology” is not something employers will pay for!

So I don’t need ontologies. I need well-documented linked-data tools that I can use and teach. I need linked-data workflows, based on real-world problems and real-world solutions, that I can demonstrate and imitate. I need linked-data systems that do real library work, right out of the box. Very little of this exists today because too much of the linked-data community is off in corners having arcane discussions about owl:sameAs and HTTP-range-14—just like XML namespaces back in the day. And I’m saying, stop that. Before you write one more line of OWL or RDF Schema, write code that lets real live people do real-world work with linked data.

Challenge #2: It’s not about what you can do with linked data. It’s about what I can do with linked data.

When I was running institutional repositories, I went to conferences about them, as you do. And at those conferences I saw a lot of demos of new and innovative software hacks. And a lot of those demos were absolutely amazing—but they were completely irrelevant to me, because they were impossible to implement in my environment. So I challenge everyone here, because you are all experts already, to stop thinking about what you can do with linked data and instead think about what I can do with linked data. And what my learners can do. And what catalogers and metadata librarians and digital-library managers and institutional-repository managers and reference librarians can do! Because if you are the only one who can do what you do with linked data, librarianship writ large will never be able to do it. And if you think this is a stealth demand for better tool usability, you’re absolutely right, it is! But that’s not all it is.

This means that you need to learn about what I do, and what I can do. And what catalogers and metadata librarians and all the rest of us do. Maybe actually watching us do it? Maybe doing some of it yourselves? Yes. So I challenge you to be curious about my work environment as an educator. And catalogers’ work environments. And digital-library work environments. Find out about those, firsthand, and use what you learn to build linked-data systems that all librarians and libraries benefit from.

Challenge #3: Wow me with linked data. Wow librarianship with linked data!

My third challenge, and I’m quite hopeful about this one, actually—make me say “wow!” about something you did with linked data. And why stop at me? I challenge you to wow all of librarianship with linked data!

Some of you may remember the rollout of the Endeca-based library catalog at North Carolina State University in the mid-2000s. For those of you who don’t recall, it was this one catalog that started the whole discovery-layer movement. What I remember most about that was that the new catalog got basically zero pushback from librarianship generally, even though it was a huge change where you’d normally expect a lot of negative path dependence to kick in. Instead, everybody said “wow.” Wow, I want that! Wow, look, facets for narrowing searches! Wow, check it out, you can actually start a query by drilling down through subject headings! Wow, de-duplicated records! Wow, relevance ranking! It was just a giant leap forward from what we had. Forget negative path dependence, people wanted this functionality now.

I challenge you to make something for libraries with linked data that has as much wow as that original Endeca catalog did, so much wow that nobody even argues about linked data because everybody wants what it can do.

Challenge #4: Disrupt MARC with linked data

Okay, I’m just going to say this: If we want MARC dead—and we do—we’ll have to kill it ourselves and eat the evidence. But I have a different idea about how to do this than I think most librarians in the linked-data space do. I see linked-data effort focusing on big national libraries, big academic libraries, big consortia, nothing but big-big-big. I’m not sure that’s the right strategy all by itself, to be honest. I’m sorry for using the word “disrupt” because I know it’s a giant cliché now, but I’m serious about it. Let me explain what I mean.

Last summer I taught another continuing-education course for public librarians, about acquiring books from independent publishers and people who self-publish. And one of my learners, who is a public librarian in a small-town public library, said a very sad thing. There was no way her library would be able to buy indie or self-published books, not print and not electronic. Just no way. Why not, I asked. Because there are only two employees at that library, she said, so they can’t do any original cataloging.

That librarian and her little tiny two-person library? They’re what disruption theory calls an “underserved market.” MARC is no good for her—it’s too complicated and too expensive. If you can make a simple linked-data system that’s cheaper and easier and more convenient for her, and lets her put in all the books she wants, including indie books, and lets her patrons find all the books they want, she will use it. So will a lot of little tiny libraries that just can’t do MARC. And if linked data is so easy and so great that little tiny libraries with two employees use it, what’s everybody else’s excuse, right? If linked data starts small, it can take over the world from MARC! I really believe this!

So if you say linked data is so much better than MARC, I’m saying prove it, for great justice! Okay, okay, last nerd joke, I promise, but the serious point behind the joke is that there really is a social justice issue here. Linked data shouldn’t be something that only helps big libraries and their librarians. Let’s build small first, and build up from there, and then we can help all libraries, all librarians, and all library patrons. I think a linked-data catalog that small libraries and their librarians can actually use and is demonstrably better than what they have can be built. Right now, today, it can be built. I challenge you to build it, for great justice—including justice within librarianship for linked data.

So once again, thanks for having me, and I look forward to the rest of the conference!