Podcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Youtube Music | RSS

Since the semantic web was introduced almost 25 years ago, many have dismissed it as a failure.
Charles Ivie shows that the RDF standard and the knowledge-representation technology built on it have actually been quite successful.
More than half of the world’s web pages now share semantic annotations and the widespread adoption of knowledge graphs in enterprises and media companies is only growing as enterprise AI architectures mature.
We talked about:
- his long work history in the knowledge graph world
- his observation that the semantic web is “the most catastrophically successful thing which people have called a failure”
- some of the measures of the success of the semantic web: ubiquitous RDF annotations in web pages, numerous knowledge graph deployments in big enterprises and media companies, etc.
- the long history of knowledge representation
- the role of RDF as a Rosetta Stone between human knowledge and computing capabilities
- how the abstraction that RDF permits helps connect different views of knowledge within a domain
- the need to scope any ontology in a specific domain
- the role of upper ontologies
- his transition from computer science and software engineering to semantic web technologies
- the fundamental role of knowledge representation tech – to help humans communicate information, to innovate, and to solve problems
- how semantic modeling’s focus on humans working things out leads to better solutions than tech-driven approaches
- his desire to start a conversation around the fundamental upper principles of ontology design and semantic modeling, and his hypothesis that it might look something like a network of taxonomies
Charles’ bio
Charles Ivie is a Senior Graph Architect with the Amazon Neptune team at Amazon Web Services (AWS). With over 15 years of experience in the knowledge graph community, he has been instrumental in designing, leading, and implementing graph solutions across various industries.
Connect with Charles online
Video
Here’s the video version of our conversation:
Podcast intro transcript
This is the Knowledge Graph Insights podcast, episode number 31. Since the concept of the semantic web was introduced almost 25 years ago, many have dismissed it as a failure. Charles Ivie points out that it’s actually been a rousing success. From the ubiquitous presence of RDF annotations in web pages to the mass adoption of knowledge graphs in enterprises and media companies, the semantic web has been here all along and only continues to grow as more companies discover the benefits of knowledge-representation technology.
Interview transcript
Larry:
Hi everyone. Welcome to episode number 31 of the Knowledge Graph Insights Podcast. I am really happy today to welcome to the show Charles Ivie. Charles is currently a senior graph architect at Amazon’s Neptune department. He’s been in the graph community for years, worked at the BBC, ran his own consultancies, worked at places like The Telegraph and The Financial Times and places you’ve heard of. So welcome Charles. Tell the folks a little bit more about what you’re up to these days.
Charles:
Sure. Thanks. Thanks, Larry. Very grateful to be invited on, so thank you for that. And what have I been up to? Yeah, I’ve been about in the graph industry for about 14 years or something like that now. And these days I am working with the Amazon Neptune team doing everything I can to help people become more successful with their graph implementations and with their projects. And I like to talk at conferences and join things like this and write as much as I can. And occasionally they let me loose on some code too. So that’s kind of what I’m up to these days.
Larry:
Nice. Because you have a background as a software engineer and we will talk more about that later because I think that’s really relevant to a lot of what we’ll talk about. But the reason I wanted to have you on, I caught a video, we met somewhere recently, but anyhow, I was watching a video with the OriginTrail folks, and you made this great quote in there. Somebody asked about the Semantic Web and just kind of offhandedly dismissed it like people always do when they talk about it. And you said, “Yeah, but that’s the most catastrophically successful thing which people have called a failure.” Can you elaborate on that a little bit?
Charles:
Yes. I think what it really boils down to with that is what … Well, how do you classify success or failure? These are actually very incredibly abstract terms. And what I was referring to originally was a bit of data, a statistic that over 50% of web pages that exist contain RDF. And what does that mean? That means that there are statements that are written in RDF syntax in over half web pages. And that didn’t sound particularly unsuccessful to me, and I’d struggled with these kind of statements in my mind for a while, such as why is RDF a failure? You hear this kind of thing go around. And I thought it didn’t really feel like it to me, it never really felt like it was particularly a failure. I mean, I built two businesses from it and they weren’t failures and we represented an awful lot of knowledge in that time and it felt like that knowledge was represented. So yeah, how do you quantify successful failure? Who’s taking the measurements and who’s making that argument has a lot to do with that.
Larry:
And the other thing that I shared that … I was doing a presentation shortly after I saw that talk, and I grabbed that stat because I think a year ago or something, Tony Seale pointed out to something like 40 some percent of websites had RDF annotations at that point. And I was like, wow. So it’s still growing too. So there’s that sort of growth of it. But immediately I followed that with a slide, which I noticed when I went back and watched another presentation of yours that you had done a similar slide where you just kind of list the enterprises that have done stuff with RDF, like RDF graphs, like your slide listed Facebook and Amazon and Uber and Siemens and Google. And I had included in mine, like LinkedIn’s economic graph and Netflix’s graph and JPMorganChase and Credit Suisse and Bloomberg and all these others. So it feels like a pretty successful effort in that regard too.
Charles:
Yeah, exactly. In many ways it’s those with the deepest pockets that have benefited the most from the technologies, which is a statement which you hear from time to time in the modern world, of course. That’s not that surprising. It’s those who can dedicate their time on maybe understanding a larger data landscape, seeing the value of it and joining it all together and representing it properly. Is it really any surprise that they’re the ones that have got the most value out of it? No, it’s not, basically. Yeah, I mean-
Larry:
Well, they have both the problem and the means to implement a solution. That’s kind of the idea.
Charles:
Yeah. Right. Exactly. And if they find that they don’t like whatever tooling they find that can support what they’re doing, they have the capability to build things themselves or join things together and buy things that maybe others couldn’t afford and this sort of thing. So especially in the early days when things are very much cutting edge and new technology sets, you don’t necessarily have big swaths of open source technology stacks available to you and stuff like that. These things are growing all the time, of course. But if we are speaking about it from a success is count metric of implementation nodes sense, then those things will have a big bearing on that. But I’m not even necessarily sure that that’s the right metric to quantify success for.
Larry:
Because we were talking before we went on the air too, about what is the right metric because you said that, okay, which portion of the data in the world is represented as RDF in a graph someplace. That’s pretty low. It’s still mostly in SQL databases, but if the criterion was like, how do we understand all that data, you can make a case for RDF. Is that right?
Charles:
Yeah, I think that’s a very good point. I don’t even think we necessarily talked about it in exactly the way that we’re about to talk about it, I think. But you are right. I mean, okay, so if what we are going to say is that the success of this is measured on how much data do we understand, that could be understood with no previous understanding apart from how to follow ontological models, then RDF is a massive success because this is the only one that really works. So on that very metric, all the others have a kind of success rate of zero to a certain extent because there was never really any overriding standardized ontological modeling concept which filtered down into it.
Charles:
Now, you could argue that maybe for example, entity relationship diagrams tried to replicate that sort of thing. So creating conceptual things which are related to other conceptual things, but they were always really created in a form that was supposed to be implemented into maybe a relational database or something like that. They tended to be bound to technology to a certain extent. So if that was an interesting metric, then maybe RDF is vastly successful because it could represent stuff in a way that’s really well understood immediately. Yeah.
Larry:
The way you just said that too reminds me that the notion of a technology being bound to the success. You could even argue that RDF just happens to be a thing that works now, but that’s not necessarily the only way you could represent knowledge. Have you given much thought to other ways or why RDF is successful in that regard?
Charles:
Yeah, sure. We’ve been recording knowledge for a very, very long time. And I mean that as in we as in people. First of all, people were recording knowledge by basically telling stories to one another and saying things, and that kind of worked to a reasonable degree. And then people started writing things down. In many ways, what we are doing now with computer science and data is we are trying to find new ways of writing down knowledge so that we can understand them as well as machines can, and the machines can understand the data as well as we can and kind of make things communicable, which in many ways is an evolution and just a new manifestation of the same thing we’ve been trying to do forever.
Charles:
So if we go way, way back then yeah, we pass stories into one another and you lose some meaning and you gain some, and over a period of time and over lots of people, things change remarkably, and hallucinations are vast. And then people write things down and then people get opinionated and people write things down in different ways, but at least then they’re written down. But in order to write them down, you need a vocabulary and you need a language that can be understood and you need translation and all these things. And then that happens over time. And then we come up with computers and then we go, “Well, how are we going to record some data of the computers? Well, we need to be able to get computers to understand it. Well, let’s use some ones and zeros or something because what do we have but switches?” Okay, well that’s great, but that’s not really much good for human consumption, even if we can calculate some things with some seriously clever computer scientists.
Charles:
So we sort of translate and abstract and translate and codify until we get to a point where we think we’ve done a great job because we can now get specialist people called database architects and analysts who can look at real world things and then turn them into tables and rows and columns and foreign keys, and let’s make up some things to make those translatable into something that an application can translate into something that a human can just about translate in the end if they’re okay with using an Excel spreadsheet. But all we’re really doing at the end is we’re trying to work out ways of being able to communicate information and data between people in a more accessible and extensive way, and this is all sort of leading to it.
Charles:
So where does RDF fit in that picture? I think RDF fits in that picture because it’s one of the first manifestations of an attempt to create a vocabulary language that fits between a machine understanding knowledge in the same structural network driven way that we understand knowledge and that the universe exists, which is things are related to things. The relationships between things, the joins between elements in molecules, the relationships between everything are essentially the most important thing which exists. And they are fundamental to graphs and they’re fundamental to our knowledge and they’re first class systems. I’m rambling a bit, but I think you get the point. So RDF is in many ways one of the first attempts to make a vocabulary, a real vocabulary glossary, Rosetta Stone, if you like, to be able to get things to translate between these things.
Larry:
As you were talking about that, I’m thinking about the graph representation. That’s just how humans think. You draw little boxes and arrows and you connect things, and whether it’s, I don’t know, football plays or relational database schemes, whatever. So there’s that sort of graph part of it, but then there’s the human agreement ensconcement in technology part of it, which is what RDF is so good at. How do you see the relationship between the meaning and the structure stuff that’s going on in this milieu?
Charles:
What do you mean by meaning and … Could you clarify what you mean by-
Larry:
Yeah, so what I mean is there’s both … As you were talking, it occurred to me that there’s something about the graph structure that’s uniquely well-suited to helping people understand each other and share things with each other and with computers. But that’s just like the latticework. And then what do the things in there mean? What are the things you understand about that stuff that stitch them together, that give them meaning? And when I think about that, I think about the semantic spectrum from just list of terms to taxonomies, to thesauri, to full-blown ontologies, just all the ways you can attempt to ascribe meaning to stuff, that relationship between the meaning attached to things and how you organize them physically in a structure. Does that make sense?
Charles:
Are you almost talking in terms of actual things and conceptual things? Are you talking about that?
Larry:
Yeah, it’s kind of like, I think in database parlance, it might be that connection between the conceptual modeling and then getting into the logical and physical modeling that you do, and how RDF fits in that scheme of things.
Charles:
Yeah. Well, I guess every layer of information is related to every other layer of information, and the lines are kind of blurred and abstract in many ways. And so what do I mean by that? RDF provides us with total flexibility in terms of an ontological modeling. And indeed, the way we describe the world, we have total flexibility in how we do it. So when we’re speaking in natural terms, we use hierarchies and abstractions in the way we speak all the time. We don’t necessarily talk about individuals. We talk about things which are higher level all the time, but in many ways, everything which exists in the universe is only the pinnacle of one other hierarchy. Always things go together to make a thing. Everything is made of other things unless you get so small, well, we can’t even work out what the hell they are anyway.
Charles:
So everything in a way is kind of a network of hierarchies, a network of taxonomies anyway. And our interpretation is almost at what level do we interpret that with it? At what level do we care about the interpretation of it? So in my application, do I care about listing car manufacturers? Well, that’s maybe the level that I want to interact with it. So maybe my ontological model cares about vehicles and cars, and the data which I’m showing to my users is the Ford, the Maserati, and the Ferrari, for example. But if I’m interested, and that’s maybe where my ontology exists, that’s where … The organization of the structure of the network of hierarchies exists there. But if I am a machinist trying to build engines for a Maserati, quite frankly, I don’t care about which manufacturers there are. I’m just working at the level of a machinist in a engine of a Maserati.
Charles:
So my view into the taxonomy, my first port of landing in the hierarchical structure of knowledge that I care about, where I set my terminology boundaries is engine part for this part of the engine, engine part for that engine, engine part for this part. And they go together and there’s 15 different types of those and those are my instances. The way that we interact with our data is very much based on perception, which I guess is quite closely linked to whether or not you could decide something is a success or a failure very fundamentally, isn’t it?
Larry:
Yeah. As you’re talking about that, it’s blowing my mind a little bit because typically in ontology, there’s this kind of domain level concern, like that’s what you’re modeling is at a domain level, and everything you just described is in that domain of cars, but the access point to it, and I’ve heard some people talk about the different kind of views into a domain, like a pet owner versus a veterinarian versus a pet store owner. But what you’re talking about is sort of like, there’s both those kind of a higher level concern but once you’re in a car manufacturing company, you’ve already entered one of those levels of concern. You’re kind of going deeper into, okay, as the mechanic, I’m interested in this. As the engineer, I’m interested in that.
Charles:
Well, right, because otherwise it’s just not my domain.
Larry:
Yeah, exactly. And I guess that’s sort of another benefit of the graph structure, is that ability to connect domains, have subgraphs and supergraphs, or however you conceive of it.
Charles:
I think this is one of the reasons why people have always found it difficult to create one of these all-encompassing ontologies. When people talk about an ontology, they say, “Oh, you know what we really need? We need to model everything. We need a model that everyone can use for everything.” It’s like, but what’s your domain when you’re doing that? So when I create this ontological model of the world, I’m only doing that from my perspective. So that’s not much good unless people are looking at it from my perspective. So it’s probably very useful. So maybe the most successful one of those will be the one where you are the most average kind of consumer looking at the most generic things across the most things. But most people specialize in something, and that’s where the value comes from. So maybe that’s one of the reasons why some of these mass ontologies haven’t been quite so well adopted as the smaller ones.
Charles:
But when you get these smaller, more granular things, which can be used in many places, things which describe much more fundamental principles such as hierarchy or time, these are principles which normally aren’t our domain. So if I was a astrophysicist, and my domain really is time and space, which I’m not even sure this is what an astrophysicist does, but anyway, you get my point. If my domain is space and time, I would suggest I’m probably not just using PROV-O because that is my domain. But if I care about it to an abstract sense enough, then PROV-O is perfectly acceptable to record when actions happened and what happened over a period of time. So yeah, I think that’s one of the reasons why the smaller ontologies that represent these domains where your abstract level into it, your view into it is perfectly suited to that, and it’s much more likely to be at that abstraction level. So yeah.
Larry:
I was picturing as you were talking too, I was picturing in addition to levels, almost like a matrix or something like that because there’s the levels of granularity or specificity or just in any hierarchy, and then the areas of concern and interest and view on that as kind of another label to it. But I don’t know. That’s a different take on it. But what you said earlier does kind of hint at something like an uber/upper ontology somewhere along the line, because you hinted at any concept can always come down, descend from another concept. But like you also said that none of these upper ontologies have really … I mean, how many of them do you feel are really adopted like the Dublin Core or gist from Semantic Arts or there’s a bunch of them out there. Do you think any of them come close to getting pretty high up?
Charles:
Well, again, I think it’s a question of your perception of the success of such things. You often see data type properties or annotation properties or whatever, which come from those ontologies in lots of other ontologies. You often see these things used from Dublin Core or from gist or whatever. Does that mean that people are representing everything they have as classes that only exist in gist and business objects and things like this? Or if you are using a SKOS taxonomy as part of what you’re doing, is everything a SKOS concept? No, not necessarily, but they all play a part sometimes. They often play a part in many places, don’t they? So I think they’re definitely not binary answers. So I think they’re all valuable pieces of the puzzle. I’d never want to say that any of these things don’t have … It’s a failure because not everybody uses it as a upper ontology for everything. I just don’t think that’s true. I don’t think that’s true of anything. There are some pieces of it that will be very valuable and some pieces that won’t, and we’ll work out what they are over time.
Larry:
As we’re talking about this, I’m reminded of why every ontologist in the world really concerns themselves with domain constraints and use cases because it’s like otherwise, you’re just off trying to figure out the whole world, which is probably … If you’re doing this for a living, that’s probably a good way to approach it. But hey, I wanted to visit another thing because we were talking before we went on also about, you just kind of mentioned it in passing in your intro, but you’ve done a lot of software engineering and were trained as a computer scientist. And I’m curious about in your original education or the way, kind of classical non-semantic software engineers, because I meet a lot of people who they kind just have that top-level flippant dismissal of the Semantic Web as this failure and RDF, what’s that all about? But you’ve kind of spanned both worlds.
Larry:
You mentioned earlier when we were talking before we went on that you converted to RDF years ago. Not that I have really a strong agenda in this podcast, but one of the things I really like to try to do is to help people who are new to this, and there are going to be a lot of new people coming with the increased interest in knowledge representation with the arrival of LLMs and stuff. Can you talk a little bit about that relationship between conventional computer science and software engineering and RDF and some of the transitions that you had to make and others may need to figure out to practice this stuff.
Charles:
Yeah, yeah, yeah, sure. Well, yeah. I was working as a computer scientist for many years, and I did various different engineering jobs and really enjoyed it and had no problem with it. And I was always working within the boundaries of what were the right things to do, various different languages, but mainly object-orientated things. And data was in relational databases mainly with some key value stores and this sort of thing and that sort of thing. So I was always very much working within the boundaries to what the technologies could do for me.
Charles:
But I think the thing which seemed to become … The profound difference for me with working with graph and RDF when I started doing that at the BBC, was that it always felt that I was working towards the computer’s rules. Whereas with RDF, it’s almost like I’m getting the computer to work to my rules. It’s data-driven to a certain extent where I’m defining the architecture of the data now and I’m not being dictated to. So it was easy to convert me maybe because I was too much of a control freak of actually wanting to make it work the way I wanted to work. I don’t know.
Larry:
Well, it seems that your control freakness, you were sort of a proxy for humanity, it sounds like.
Charles:
Yeah, exactly. Yeah, you’re quite right. That’s a terrible turn of phrase to use for myself. But in many ways we are fundamentally trying to use computers to help us to communicate, to communicate information, to work out new things, to solve problems, to share data, all of these things. And with all of the traditional ways that we worked with data, it was kind of difficult to do that. This was the first manifestation that I’d seen when the data really was the most important first class citizen of what we’re doing, or shall I say, not the data. That’s wrong. The understanding, the meaning, the domain that I care about, for example.
Charles:
And sure, there were other attempts at domain-driven design and stuff. I did some stuff which was, okay, well, let’s think about the data structures we’re building first, but still, then you were still constrained to things which weren’t natural. So things which weren’t just describing the world. With RDF and with this stuff, I guess the reason why it was so easy to convert me is because it felt like I was actually speaking again. I was speaking to someone who spoke the same kind of conceptual language, who understood the universe a bit more than trying to translate for something which needed it to be just so, just so that I can work with it. So it has in many ways, felt like a sort of conversion or a religion. I think for a lot of us in the industry actually, we think like that.
Larry:
And also, as you just said that, it reminded of something you said earlier about the original way we shared, we did all these tasks that you were just talking about, how to communicate and share new things, share data about the world was just talking, back before even writing. And the way you just said that, felt like I was actually speaking again. It’s like, wow, I’m back to my roots as a human being sitting around a campfire telling my friends about how to do something or how to get something or yeah.
Charles:
Well, that’s right. I mean, people were designing ontologies around campfires long before we were doing it on white boards or blackboards or anything like that. Philosophers, well, some of the most interesting philosophers as far as I can tell wrote things a hell of a long time ago, and that’s basically what they were doing. It is surprising as well. It’s a really enjoyable task. If you are a computer scientist coming into the world and your first job is to go and build an application and you are doing it in, shall we say, the more traditional computer science way. It’s incredibly sort of computer science-focused learning. It’s like, “Okay, so we’re going to use a database and which database should we choose? Oh, we’ll choose this one because it’s … Well, that’s the one we’ve paid for. It’s a relational thing and it does this.”
Charles:
“Oh, right, okay, so how do we do that?”
Charles:
“Well, we’re going to send you a training course, and it’s called Relational Databases 101, and you’re going to learn all about how you join these things together. But this is a so-and-so kind of database. So you’ll need to write these things called stored procedures for this thing because … Oh, and we need to do this.”
Charles:
Okay, great. And then about six months later, you start writing your application and you’re fully focused because you need to make everything work exactly the way the application needs it to work. That’s the way you kind of used to it. But then when you’re building an application for using Semantic Web technologies, you go and sit around a campfire and discuss the meaning of what you’re trying to build. It’s just way more interesting and way more rewarding than the alternative. So if you want to bring people into this … If you find it as rewarding as I do, then it’s actually quite easy for people to get into doing it because that’s really what it is.
Charles:
You go and sit around a campfire and you discuss how you’re going to solve this problem, and you draw it on one of the tooling sets, which is available, or you just draw it on a whiteboard. With the knowledge that you know that the translation, that you know that the first versions of these vocabularies and these glossaries and these translation kits are there to put it into a machine for you. You don’t have to think like a machine anymore. You can think like a person around a campfire. You can be way more inventive and creative about what you’re going to create. And by the way, it’s much easier, much easier to come up with new ideas of new wonderful things you can do with your abundance of data if you’re sitting around a campfire riffing about it than you are if you’re stuck in the weeds of how to make this particular thing record in this particular way so that this particular other downstream process can even understand it.
Larry:
I love this so much, the way you’re describing this, because it’s like the notion of knowledge representation. I mean, knowledge is things that human beings agree on, and it’s almost impossible to do that in the scenario you just described. Like, “Well, first we have to learn this whole tech stack and blah, blah, blah.” Whereas RDF is like, “Well, yeah, you just need to know RDF so you can take those dots and arrows, you just connect it and put them in a computer.” It’s pretty simple. Hey, Charles-
Charles:
Yeah.
Larry:
Sorry, go ahead.
Charles:
No, please go on.
Larry:
No, I was just going to say we’re coming up close to time, but hey, before we wrap up, is there anything last, anything that you want to revisit from the conversation or you just want to make sure we share with folks before we wrap up?
Charles:
I think there’s just one thing which we spoke about a little bit, which I’d love to know if people have any opinion on maybe in the chat afterwards or whatever, or could come to us independently, which is we were discussing if you were going to describe the perfect way of building ontology, what is the most fundamental sort of upper principles of ontological or modeling, should we say that we can apply to the world? What could it look like?
Charles:
And I think before this, we were describing that maybe it could be something like a network of taxonomies. And I’m interested in exploring the idea of how do we really model everything? And I think it is something like your view into a network of ontology, at what layer or what dimension is your view into a network of taxonomies? Because everything, to me seems like it is at some point part of a taxonomy. And whether or not that’s an instance or whether or not that’s a real thing you care about or if it’s a conceptual thing is depending on your perspective. So I think that’s one thing that I’d like to discuss another time with you, Larry, I guess is just, yeah.
Larry:
Now that’s fantastic. And one thing I’ve thought about with this series is trying to make it more conversational because this is just, we’re having fun and this is a great conversation but I would love this to be a broader conversation. I’ll put my thinking cap on and try to figure out how we might open up that because I think there’s a lot of people who would be interested in and tremendous contributors to a conversation like that. Hey, one very last thing, Charles, if folks want to connect with you or follow you online, what’s the best place to find you?
Charles:
Yeah, you can just find me on LinkedIn. That’s probably the easiest thing to do. Charles Ivie, you’ll probably see my name in the title of this podcast or whatever. Easy to find. Connect to me on there, send me a message. I’m always eager to talk about this stuff.
Larry:
Yeah, well, I’ve had a blast talking with you about it. I’ll say that, and I’ll put that in the show notes as well. Well, thank you so much, Charles. This was a blast.
Charles:
Thank you very much, Larry, for having me on. Really appreciate it. Cheers.