IA Summit 2015 Main Conference Talk
Topic(s): metadata, ontology, and semantic web
Information Architects codify information’s purpose and context–that is vital for effective use and management. This talk summarizes our approach to understanding dynamic, multi-dimensional information models and the practical tools that help us create them.
Acquisition of heterogeneous content and data sets is growing exponentially; in many cases outstripping an organization’s ability to digest it, share it, and apply it. IAs are the ones to bring structure to all types of data, as we creatively balance descriptive integrity, simplicity, and agreement/disagreement. IAs need to be actively involved defining the logic underpinning information, to help link data externally with the semantic web. Technology tools are often sold as “magic”–claiming algorithms will automagically catalog information and surface meaning. Or content owners assume anyone (in the organization or the public) can easily apply structure to information.The skills that IAs use to help people understand content and data also help computers make connections between the data as well.
We encourage IAs to commit to models, semantic patterns and metadata. In our “Modern IA Manifesto,” we encourage you to grasp the key techniques and tools in the IA’s arsenal to provide more sustainable structure and thus provide more flexibility to information.
About the speaker(s)
Dalia Levine is the ontologist at HBO. Her focus is creating consistency in metadata across the enterprise. She works on classifying information across multiple sources, documents, and media. She contributes to and incorporates international standards to streamline workflow and capture knowledge. When she organizes information, she helps to get the computer to understand what you mean. She was the co-chair of the 2016 IA summit and is always trying to think of good questions.
Duane Degler is a partner in Design for Context, a Washington DC/Baltimore usable design consultancy. He specializes in the design of sophisticated interactive applications and search experiences, making rich data resources usable and relevant. Duane is practice lead for IA and Cultural Heritage. He has led projects for commercial and government clients in the US and Europe. He authored the “Dynamic IA” chapter in the book Reframing Information Architecture (2014). Since 2003, he has focused on the unique challenges and opportunities that arise when designing specifically for linked data and semantically-enabled applications.
Dalia Levine: Welcome to Structure and Metadata: Shortening the Onramp to Linked Data. I am Dalia Levine. According to Ted Nelson yesterday, I am Madame Librarian. I am an Information Architect. I work with metadata and I’m here to tell you that you should as well.
Duane Degler: I’m Duane Degler. I’ve been working in Information Architecture and Interaction Design for many different types of systems for a long time. Fell backwards into the semantic web and linked data about 13 or 14 years ago, and wonder to this day why is it still so hard to do. What we’re going to do today is talk through not the details of what you do in here, but why we in this community should own this. Why we in this community really should be taking all that we know and all that we work with and all that we do in looking at metadata and structure and information in order to shorten that onramp.
When Dalia and I began talking about this about a year ago what we were trying to figure out is what’s the greatest point of failure for organizations to adopt all of these cool data models and technologies that are going to really make the web and our internal systems that much richer and easier to use and more flexible to work with. The problem is that it’s a big lift.
We’ve all dealt with the metadata problem of people saying we’ve got metadata but nobody every fills the fields in. The keywords never mean anything. It’s too hard to build indexes around what we’re doing so nobody can ever find anything. We did this really big taxonomy project and then we did another one and then we did another one, which is kind of like it sank so we built another one and it sank.
This is going to happen. It’s going to happen with this group of people and people like us out there in the world.
Let’s start by saying what is it that’s the scope of what we’re talking about here. When we talk about things like metadata we are talking about the basic core of metadata inside of content and within systems that we work in which is the stuff about an “it” itself. An item of data, document, content, whatever it is. The stuff about it.
We’re really talking also and primarily about how that thing exists in the world. Typically, we use things like subject keywords as ways of saying, “Here’s what this thing is about,” but that’s often a very limited and blunt tool for what something is about. Then, there’s the other aspect of it, which is how we think about any one thing’s relationship to other things. We’ve got a lot of stuff out there. There’s stuff everywhere, and all of that stuff is floating in little isolated islands and silos.
These relationships that we build between things become incredibly important. When we think about the idea of Aboutness. Aboutness is more than a pile of keywords. It’s more than saying, “Yep, if we populate 10 strings of text and put commas between them, we’re good.” We can do that, and we can even do that and say, “Well, subjects, nouns…” but when we link things together, it’s powerful.
We know that from the web, we heard that from Ted last night, we’ve known it for years, that’s what we do, but the power in the link data piece of that is naming the lines. It’s giving verbs to the web. Jorge yesterday morning said, “Let’s think about structure in a more disciplined way.” We can start by thinking about creating statements and sentences that mean something, expressing something with a statement.
Dalia: There have been a lot of talks mentioning ontologies. There have been talks mentioning marking up the data field. Do you sense a theme? We’re going to keep this theme going.
Duane: We’re also going to use jargon like ontologies. Very quickly, when we talk about these disciplined structures, we’re talking about things that we know, things that we can express, things that have identity, maybe a person or an organization. There may be a binding role between them. Rather than just saying, “This person links to this organization” or vice versa, there’s actually a role that that person plays in relation to that organization.
By naming the lines, we can assess that role. We can understand the role that plays. You can get different kinds of information, potentially from different sources, and by itself, if they’re just baskets of keywords, if they’re just piles of keywords, search engines can go and chew that up and use that, but again, the power comes from starting to link these things and describing the nature of the relationship between things.
It’s very different if I’m saying, “I’m going for a beer with you,” “I’m lending money to you.” Totally different contexts and meanings that come in the naming of the predicate relationship line.
These things, at the higher level, above the individual instance of me, but me as a person, are ontologies. They’re essentially a structure that has been given to a set of concepts and their relationships between each other.
Anytime you look at ontology, that’s what we’re talking about in computer science terms. We’re talking about a structure of a set of concepts and their relationships with each other. It’s a very, very powerful thing to have available to us.
Why are we talking about this now? Well, we’ve got a massive deluge of content and data and everything else out there. Search engines, as we can see, are no longer sufficient in and of themselves for people to actually use to navigate and manage, whether we’re thinking about the Deep Web, whether we’re thinking about the huge volumes of content.
Anybody who’s dealt at all in enterprise search and intranets knows how much people struggle with getting the things they need in timely manners without having to churn.
A couple years ago, everybody was saying, “Don’t worry about it. This deluge, big data, no problem. We’ve got the AI tools, we’re going to nail this one.” Algorithms alone are not the answer to that, because machines by themselves can’t interpret everything that needs to be interpreted to understand the human need for the meaning.
Dalia: What do we mean? Most of our time is spent building web pages. Here’s a web page from a Smithsonian museum. This looks really familiar. There are labels, there are categories on the left, it’s navigation with some words.
Nod heads, this looks very familiar. This is what we deal with most of the time every day. This is our lives and what we’re building.
Occasionally, there’ll be a link, John Singer Sargent with this hypertext, and maybe we can go somewhere else. But if a machine is reading it and the machine is parsing it, and there’s no meaning behind it, machine just sees blah, blah, blah, blah, blah, and there’s no meaning.
Duane: Again, the algorithms and the folks who will build AI tools will tell you, “But that’s all right. We can parse out all that text. We can build sentence structures and understand these things, and it’ll be automagic.” Not even automatic, but automagic.
These tools have gotten a lot better in the last few years. There is a tremendous amount of things that we can do now. There’s a tremendous amount of leverage that we can get in managing meta data from these automated tools, whether it’s text mining, concept extraction, various types of semantic enrichment, machine learning, pattern modeling of dozens of varieties.
It’s an explosion right now of tool sets that are becoming available to us. Very intense.
There’s also a tremendous capability. If we think back to XML and how hard it was to sort of manually craft things, if we go back to the origin of HTML and how many of us were hand coding pointy brackets, we don’t really have to do that kind of stuff even with the link data anymore.
How many of you are familiar with or have worked with Drupal? Drupal 7 or 8? A fair proportion of you. The RDF link data modules are baked into the product now. You can just turn them on and metadata that you put into Drupal will be exposed out to the machine-readable web in these ways, as link data.
It’s not quite a magic wand, it’s not quite automagic. There are things you want to do for this, but it’s there. These things are being baked into our tools every day. We now have to begin to know that and use them.
Visualizations, getting better and more powerful all the time. It’s probably going to be the biggest legacy that comes out of big data. It’s all of the flexible visualization tools that we have available to help people interpret and understand really, really large scale data sets. It’s really more possible than ever before.
Now, one of the things that we hear a lot in our practices and when we’re working with people is, “Oh, but we’ve got all this legacy stuff and we can’t really deal with it.” Well, we would argue that link data is really the next generation data standard. This isn’t, in its own sweet way, a revolution. At least that’s what we will tell the IT department.
It is an evolution. It’s a natural evolution.
If we think back to where a lot of us have come from, card catalogs. What are card catalogs about? Card catalogs are about an organizing scheme and a consensus that people had around an organizing scheme to take large volumes of complex data and provide an entry point to them. It was really about establishing organization in a shared way.
The second is, then, as we introduce more in computing, we see more relational databases come in. We can take and translate those organizing schemes into structured things that machines can read and use and do processes on, run math on them.
Now, with the proliferation of electronic documents and then, later, the proliferation of the web, we see the emergence of XML. What XML is doing is is creating an independence from the specific formats and varieties of things that we were seeing in relational databases.
It makes it easier and more accessible to share things with each other, so it’s independent of the individual technology. It’s more persistent that way.
RDF is resource description framework. It’s the atomic element of link data. It’s really at the heart of where link data is. It’s the first step in there.
You’ll see that O word again, ontologies.
What problem is this trying to solve? The fundamental problem it’s solving is the problem of expressiveness, because all of these others are structures. They’re structures that are not necessarily as rich or expressive for carrying meaning and helping people with understanding. That’s why we think it’s pretty powerful for this community.
Dalia: Information architects, we know these tools exist. But we’re very good at the language, we’re very good at the verbs. The verbs are the expressiveness.
This is what we need to own. We need to be, not just a passenger along for the ride, but in the driver’s seat. In order to get to that, we have a manifesto for information architects on how to get to link data. We have five points. We’ll take you through each one of them.
The first one is build on what exists. There’s a lot of information already out there. You don’t have to start from scratch.
Now, you could start from what is already out there and then add your own parts as you need to. But the more that you can link the data, the more that you can find it and express the meaning, is start from what’s already been built. It’s 2015 and there have been a lot of vocabularies, a lot of ontologies already built. They’re out there for a reusing. Go ahead and reuse them.
There are standards for specific domains and there are models that already exist, so you’re not alone in just starting off with effort. There are things to tap, there are places to go.
It’s really hard to diagram all the vocabularies out there. This is just one diagram of the vocabularies. This is from the Open Knowledge Foundation.
There are people vocabularies, there are geography vocabularies, there are food, government, environment. As you look about, you’re modeling your information and you’re architecting your information, you can tap what’s already out there and use it to express the meaning that you’re trying to use.
Duane: While we’re talking about these vocabularies, the ontologies that you can use, recognize that there are different types and different flavors that we can use in different ways. As you see them, think about what suits your needs.
On the vocabulary side, as Dalia just said, there are the common available vocabularies, there are numerous of them. You saw the 500 on the previous slide. There are a lot of different sites that are beginning to aggregate and provide access to those.
Then there are local vocabularies. They are those things that are unique to you and your organizations. You want those to remain unique. Maybe it’s your product catalog, maybe it’s your organizational chart. Maybe it’s aspects of the description about who your organizations are.
Then above that, there are the domain ontologies. We’re seeing, in the domain ontologies, quite a proliferation. The medical community’s been moving very rapidly in this area. There are a tremendous number of ontologies and descriptive models now for medical related things.
The financial industry. Financial industry is very, very scope on sharing information and the regulatory environments are making demands on that.
Government, if you look at all the open data.gov and data.gov.uk, and New Zealand and Australia and just about everyplace else now, those government oriented ontologies are available. If you’re trying to express something in and around that domain, there’s probably already an ontological model that you can start from to do that.
Above that, we have general ontologies. General ontologies are things that describe particular things. A person, how do I want to describe a person, and person’s relationship to people in organizations? It’s a very simple way, there are things like, friend of a friend, FOAF.
How do we build thesauri? There’s another general ontology for actually modeling and building thesauri in the link data environment.
Geography, events, there are numerous of these general ontologies that can be applied in all sorts of different disciplines and help you actually mix data from different places and different disciplines together.
At the top level, there’s a set of sort of upper level ontologies, which are the sort of big overarching concepts, society, government, human beings, biology, these very, very broad structural concepts that sit in there, as well.
One of the key things that we have to work on as we get into individual projects is bridging ontologies. Because it’s very, very rare in this space that you’ll just use one ontology and just use that and it suits all your needs.
Very often, what you’re doing is drawing in different ones from different places and there will be points of crosswalk between them. When you think about the information architecture practice, part of that information architecture practice is going to be in that crosswalking between these different conceptual frames of reference.
When you’re building local ontologies, when you’re building your bridging ontologies, when you’re building your vocabularies, we encourage you to share them wherever possible. If it’s not totally proprietary data, and very often, the ontologies above the instance level are not proprietary, it helps other organizations come on board faster as well. There’s a greater collaboration if we can each use things from each other.
It’s essentially how the web grew up, HTMLs in exposed open environment. How many people pulled up pages, went to the source, scraped the little thing off and built themselves their own page? These are the same kinds of things. These are shareable.
Dalia: As you share the information, you increase the value of the information. It becomes richer. Even in the finance world where a lot of that information is protected for personal data information, there are parts of it that can be shared. That part is also based on what we were saying before, it’s in the middle. It’s that verb. What do we mean when we say this?
This is where we can, as IAs, really help out. I’m going to use an example that I’ve used in my work and I’m going to use GeoNames, which is one of the largest location geography. I think OpenStreetMap is based on it, so you never should ever build your own location vocabulary again.
Because if you start using the codes and the bases from GeoNames, you get the context that GeoNames already has. If you find a mistake in GeoNames, it’s edited by the entire world. Like Wikipedia, you can join in and edit if you find a mistake. I have yet to find a mistake.
Springfield. There are 1,383. Did you know there were that many Springfields? These are only the ones in North America and Central America, and if you see on the very left, there are some in Ireland and in the UK. Which Springfield are you talking about? There’s a reason why “The Simpsons” used Springfield as their location, because it could be anywhere.
On the left, and it’s very small on the slide, and you can go to GeoNames and see this yourself, there are features about each location. Is this place a stream, a state? Because it’s a collective knowledge and there’s aboutness about each one, you can tell that this Springfield is in Massachusetts and it has this longitude and this latitude.
That really helps, because we mean that Springfield and not the one in Australia, or Southern Africa, or New Zealand. Did you know that there were this many Springfields? I love this example. What does that mean? In my work I actually took Springfield out of our local version of our vocabulary because we need to disambiguate.
What was most important? I knew, with trust, that using the code or whatever I wanted to pull from GeoNames, I was referring to this Springfield. Not only was the human knowing that it was this Springfield, the computer knew this Springfield.
Later on, you could have found the different connections. That’s linked data. Now you get the data about the place, so if you ever use Springfield, go to GeoNames and use the specific Springfield.
Duane: It works for other towns and cities as well, by the way.
Duane: It’s not exclusive to Springfield.
Dalia: But you can’t make a “Simpsons” reference.
Duane: How many of our subject experts are going to want to go out and find latitude and longitude numbers and type them in to metadata fields? The answer is not many. Part of what these tools do as well in getting us to shorten that onramp to adoption is make it easier to put, for people who are not detailed technologists, to put information in in a more flexible way.
The power of this is its ability to extend out across different kinds of institutions and different kinds of domains. Here’s just an example of carrying this out into the world. We were all in Baltimore, many us, I think, were in Baltimore at the IA summit a couple of years ago.
Some of you, if you were wandering the city, may have wandered past a building that, at this point is empty, and say, “Wonder what that is.” It’s pretty easy. We’ve got phones in our pockets, we can take pictures, we can query it up, we can look at the map, and it’ll tell us, in fact, it used to be a museum. It used to be Rembrandt Peale’s museum.
It happened to have opened around the time of the War of 1812, in the finish of the War of 1812. It was actually the first place that held the flag that flew over Fort McHenry that Francis Scott Key wrote the poem about, which is now sitting in the Smithsonian. If I’m there at that point, it’s a quick day trip down, 40 miles down, quick train ride down, to go visit this thing.
Just by walking by the building, I’ve suddenly discovered, “Hey. I can go visit this thing, and while I’m there, let me go to the museums and look around in the museums and see what kinds of things are there — and, by the way, there are a few paintings that have been painted by him, both in the Smithsonian and in the National Gallery of Arts.”
From one simple starting point, we’ve move through a whole series of different cultural institutions that are all sharing their data with each other in an open way. Maybe we want to stay in Baltimore, we don’t want to take that train ride. It’s nice to have that over there but, in fact, let’s see what we can do in Baltimore.
The other thing that we can learn, and this was something that we actually did discover two years ago as we were doing this, is that the artillery memorabilia from that period was also in that museum and a lot of it was coming out of places like Patterson Park. While we were there, there was a dig, an archaeological dig.
If you wanted to get out on a Saturday and get your spade out and go have a totally unique experience in Baltimore, it was possible. How do you find out about it? Maybe the hotel concierge will tell you, or maybe you’re able to discover it by walking by a building.
These ontologies that come from different domains and cross connect with each other and that bridge with each other start giving us an incredible amount of power to express concepts that cross boundaries and bridge understandings and help with discovery.
As we think about that and carry that idea forward, what we’re thinking about here is how we can encourage, as information architects, understanding and knowledge building in a much broader way, in a much more sustainable way. When we talk about knowledge building let’s talk about the interconnected information, the example I just gave you, happening in numbers of ways.
I did a project about 18 months ago with an institute that was working for food manufacturers, a number of consortiums of different food manufacturers. This institute was taking data from various countries around the world on various different indicators.
The goal was to put on the CEOs’ desks of all of these food companies indicators that, when they went out into these countries and established either deals with raw-material food producers or manufacturers who were doing particular things, they were putting those things together.
They were actually showing economic indicators, social indicators, environmental indicators, and whether they were getting better or worse over a period of years. We’re beginning to see some real power in reconnecting and connecting different concepts. We also want to be assured that that’s happening over time.
The other thing about using these standard syntaxes and these standard open models of data representation is they’re sustainable. We work a lot in the museum community where they’re thinking in 30, 50, 100 year timescales. This is an ontology that has been built primarily through the cultural heritage community for the establishment of cultural heritage information.
Simple, right? No worries. We can do that. Give me an afternoon and a hundred thousand records, we’re on it. As complex and rich as that is, and as complex as it looks like, and this is a lot of what you see when you’re looking out…
You’re going out and you do research on linked data and you get things like this. When we unmask it, what we’re doing when we unmask it is say, “This stuff isn’t that different than what we handle today.”
Dalia: This is what I Azed you. We categorize. Here’s the collection in rights over here. Here’s the visual representation down here. Bibliographics over here. Now it makes a little bit more sense. This is what we do.
We look at complex information and make it chunks or stuff or blobs, whatever you want to say. This is IAs. Start helping out to help those really complex little ontologies behind there make more sense.
Duane: It also helps because a lot of this stuff is already sitting in databases in these forms. It’s just not in this syntactic representation and it’s not exposed out in the world. There are many people who are managing their data very diligently, in a very committed way.
It’s just not getting out of those relational databases and the more we can make it approachable so it seems that what people have here is just going here and becomes more flexible and more reusable. The goal here is to really begin to give us much more capability to foster insights, to be able to look at identity of things and disambiguate that.
That point down at the bottom, to really help us support the machine interpretation of a world view that we as humans can express. This is not about the machines and the algorithms doing all the interpretation. This is actually about us doing it and enabling the machines to help.
Dalia: The other point on the British museum is that there are other institutions who don’t have the same extensive collection as the British museum, but they’re connected through that complicated ontology.
Now this institution can refer to they have a fragment of the Rosetta stone, but the real Rosetta stone is over at the British museum. Using that ontology, connecting the cultural institutions together, now you get a collective understanding of the world view.
Duane: There are a couple of groups out there like the Information Sciences Institute in USC who built tools for mapping relational data sets into linked data. It takes them about a week to map a reasonably complex collection, and then they hit a button and every night the batch runs and does the updates.
This linked data is all there and available in a very quick manner. We know some people outside of the cultural environment who’d been doing that with EPA data and NIH data to look at environmental and health-related things.
It’s taken them about three weeks to pull data sets from National Library of Medicine, NIH, EPA, and really make sense of it. It’s not a long process once you have the tooling available.
The other power that it brings us is responding to context, which, of course, we all will agree, I hope we all will agree, if not, talk to me afterwards we’ll convince you that this is important. We have to do this.
We hear a lot about this because what we’re talking about here in context is being able to meet people where they’re needs and expectations are. That might be in a place. It might be socially with sets of people. It might be multiplatform across devices in some form or another.
We want to be able to meet people where they are and really help enable them. If we think about the way that we interact with the web now in a lot of cases, it’s one directional. We enable structures and people can go out and find things, but people are directed to go out and find things.
Where we’re headed is that profile and context models sit in the middle of that and help make that a two-way interaction between the information and the individuals. Done a number of projects on things like news streaming, recommender systems, and they all have underneath them very rich, descriptive, semantic linked models in order to be able to help people resonate.
I will say, to do that, we have an incredibly powerful tool in Linked Data, because the key to really, really good, relevant, personalized things is not just subject keywords. It’s not just, “Oh, look at that. I’m interested in armadillos and you’ve got documents that are tagged with armadillo. We’re good.” No. We want to be able to look at things like what is the user, what’s their experience, what are they trying to do, what situations are they in at the moment, what have they been looking at in the last few minutes that might be influencing the patterns of what they’re going to do now.
The cool thing about this is we can model all of those things in the systems as Linked Data. They’re not now. All of these different things are probably different silos inside your applications. They’re probably modeled in very, very different ways. They don’t have to be. Once you match that all up together, it’s incredibly powerful.
Dalia: The last point of our manifesto is my favorite part, is that you have to be serious about the data in your care. We’re still humans. Machines are not going to take over. We need to be a part of making sure that the machines have what still makes us human part of them.
The balance of the human and the machine. Humans describe the stuff. Systems capture the stuff. They might ambiently capture the metadata. Technology then identifies the stuff as all things that we’ve all talked about. We create the environment. We create the meaning making. We understand it more than the computers, more than the algorithms, more than the machines.
For me personally, I’m identified as a librarian. I’m going to let you know a secret. I don’t like cataloging at all.
Dalia: I can’t understand.
Duane: You’re not alone.
Dalia: That’s the place where the information has been captured over time, but language and the information changes over time as well. The technology lets us get at it. We get at what you mean. I know how to question technology to make sure it’s accurate, to make sure that it’s helpful, to make sure that it’s valuable, that you get the computer or whatever you want to call it or whatever it becomes called to find what you’re looking for. That’s where the human still needs to participate in that.
It’s a responsibility as well. It’s a responsibility to people. It’s a responsibility to information integrity. It’s a responsibility to meaning. It’s a responsibility to the future because the machines can’t do it without the humans, and the humans were not going to be able to deal with the influx of data that we already have.
As Karl Fast said a couple of years ago, the small data problem, the big data problem, there are still the humans. That’s why IAs need to be a part of the discussion and be a part of this.
We are the humans. To demonstrate this, I worked at the Ford Foundation and we have a board member who is described as an Afro-Columbian. She’s the former minister of culture, Paula Moreno. If you used the standard classification on the equal opportunity if you work in the United States, that you have to fill out, she doesn’t quite fit. This is going to be happening more as society shifts and changes.
If you use an ontology to describe her, you can better accurately say Paula Moreno is a lawmaker from Columbia which is in South America, and she is of African descent. You got much more meaning, much more accuracy from that sentence than if I bracketed up. You could say where the subject-predicate and those little, squiggly lines that we’ve been talking about are so important.
Here’s another part. I said South America because I was just looking at GeoNames and it’s very clear. When you’re in a legacy institution who had categorized the world at some point, not sure when, they described it as Latin America. What does that mean now?
With GeoNames connecting it, we could say our information, our local definition of Latin America is these countries. These are the GeoNames for these countries. Now, it doesn’t matter because those GeoNames are set. They’re understood. They’re understood by the humans, and they’re understood by the machines.
Now, we have a much more accurate expressive meaning that’s also connecting with the open data at the world. It expresses that without having to read the entire sentence, without having to make judgment, the accuracy and then the human part of it.
Duane: There’s a very important point in that story. There were many important points in that story. I’m not going to underwrite it. One really important point that I want to call out here is the fact that you can have multiple expressions of things. Whether something’s expressed as South America, Latin America, it doesn’t have to be a one-to-one overlap. There are going to be a bit of a Venn diagram in there.
That kind of expression, the fact that the language changes over time, the evolution of that language can be codified in ways that’s much, much harder to do in a lot of other existing systems by creating these links and these bridges between information. It’s incredibly powerful. I’ve been saying that a lot, sorry. Do you think might care?
Let’s round out here and ground this back in thinking about what is important for you as information architects. The architecture piece of that, if we think about built architecture, it’s about sweating the details. It’s about really focusing early on because it’s hard to change things later. It’s very important to make sure that we’re really deliberate about the way we think about our information, our information structures, and the way that they work.
We would argue that you come to this work in Linked Data with an architecture mindset. At the same time, this is really about information. It’s about things where there are these nuances, where there are these multiple levels of meaning and interpretation that need to be described.
Its richness is really important. There’s a whole legacy of discipline around information that we in this community have been fostering for 16 years in this venue and for many years before that and many people before us have been establishing that.
We want to, again, bring to our work an information discipline that supports that architecture mindset. We’re also in a different place. Technology is in a different place. The technologies that we’re dealing with now allow a much greater and more dynamic use of the technologies and interaction between the technology and the users.
If we think of it, there’s so much with agile and minimum viable products and permanent betas and all of that. It seems that everything is constantly changing all the time. Those first two points are about grounding in the things that really matter and making them solid but recognizing that as we heard so much yesterday, if you think about what Marsha Haverty was saying, if you think about what Andy Fitzgerald was saying, if you think about what Jorge was saying yesterday, if you think about what Tedd was saying really at the end of the day, that these things are constantly going to change and move and become more enabled.
What we don’t want to have to do is constantly recast and re-manage and rebuild and rework our data every time we do that. We need a flexible data construct underneath the covers of all of these. We care. I think you got that.
Dalia: To sum it up, just remind you where we started. This is our manifesto. Build on what exists, increase the value of the information by building on what exists before. Encourage understanding and knowledge building, they all build on each other, respond to the context, and then be serious about the data in your care. Then, you’ve got it. We’re in the driver’s seat and we’re helping move along. Whatever the future holds, we’re in this together.
Duane: It’s easy. We’ll check with you at the end of the day and see how you’re doing.
Duane: Thank you very much.
Dalia: Thank you very much.
Dalia: We have three minutes for questions.
Audience Member: Thank you so much. That was amazing and very actionable. I really enjoyed the talk. My question is on linking— I don’t know if it’s a question. I’m just wondering your thoughts on cross-connecting subjects within your own product or within your own site and cross-connecting with other organizations and other sites and strategies behind that or just any thoughts there.
Dalia: I’ll take it real quickly that at the Ford Foundation, as a major philanthropic organization, most philanthropies report to the foundation center individually and as a foundation center, then handles that data. The foundation center is actually working with all the foundations.
What I did was I grabbed the vocabulary from the foundation center. They want an input. I double checked it against ours. It’s a long process right now and this is what’s happening in why IAs need to be a part of this process. We started making sure that what they meant was what we meant. The language, the grant makers, the subject matter experts we’re using was found in the other ones. It’s a continuous dialog, but that’s also where location is where oftentimes people start off with.
We just started codifying locations so that if we knew a grant was going to Springfield, Massachusetts, it was that Springfield, Massachusetts. As we pushed out our data with the formatting for the Linked Data, the foundation center knew which Springfield, Massachusetts that grant was. Then, it’s actually out in the open now. Anyone else can get it. Now, it’s the same. Now, you can build tools between those.
We, as a foundation, kept some of the information like how we organize our grant making and this is all in the website. That was ours. That’s the language we chose, but we connected it with the subjects the foundation center was using.
It’s finding those standards that might exist which is why libraries and the history of cataloging, there have been some standards to grab on. There are others. There are industry codes. There is data from news organizations. There are a lot to choose from. Find little bits, don’t start huge. The little bits, just start with one. Then, that’s all you need. You just need one.
Duane: One quick point on that. Sorry, I know we’ve got a question here. One quick point on that is you never know when your company is going to get bought or buy another company or something else. The more you use standardized, accessible, and available things at least in your domain, the better able you’d be able to respond to changes in the business environment as you go forward because you’ll already be in a normative form.
Audience Member: Thank you. Excellent talk. I don’t know if this is a question or more of a comment from what I’ve seen. You talked about XML. Every time I bring up XML at the IA summit or around IAs, it’s more of a four-letter word than a three-letter word. It’s this big, scary, ugly thing. I come from the data architecture world where it’s all XML. For some of these standards that I’ve had to work with, one is national information exchange model.
Duane: That is a big, scary thing.
Audience Member: Adam, if you’re in here, he’s worked with it. It is ugly. When I say ugly, it’s ugly because it’s made by people that don’t have the discipline and the rigor of taking things that are user-centered design. It’s not user-friendly at all. How do we get the practice and taking a looking at using tools like XML to do some of the things that we do to actually bring it out of the data world and into the information world to get some of these rich experiences.
Dalia: See the talk next year.
Duane: That will help. What I would say to you is, as Dalia pointed out in the last question, start small.
Duane: Find a particular problem or a particular application or particular circumstance where you can do that. Also, in the design practice, we have a tendency to pair up very, very rapidly with the architects and start a dialog around the object models. Even if they’re planning to put it into relational databases or XML or anything else, the modeling language, treat the stuff as a modeling language as much as a technology.
If you think in this way of relationships, it makes it much easier to carry that forward into other technologies in the future and to get people to resonate with it at a developer level and at a human level. Sorry…
Dalia: Something really, really, really small is something like there are five different definitions of account number in one organization because of different systems and other stuff. Model that, link it to say, “If you have an account number, can we agree on whose is the standard account number? Then, the rest of you refer to that.” “Oh, we have a triple.” Start there.
Seriously, five different account numbers in two floors of one building in one company. It’s going to continue because that’s my system, that’s your system. What if these systems have to talk to each other? “Oh, great, triple.”
Duane: Leave the systems alone and abstract above it and actually do that with the semantics.
Dalia: We’ll be around at lunch for more questions. Find us anywhere. Thank you very much.