Latest / The Adults In The AI Room / Andy Fitzgerald on IA & Structured Content Design - Episode #23

Transcript

Andy Fitzgerald on IA & Structured Content Design - Episode #23 · The Adults In The AI Room

• 0:04
  SPEAKER_00: Welcome to Limited the Early AI Podcast.
• 0:07
  SPEAKER_00: Here we talk with thought leaders and practitioners on what's possible and practical with AI and the enterprise as we move from the current state into a future of cognitive assistance and embedded AI in all aspects of our personal and professional lives.
• 0:22
  SPEAKER_00: Thanks so much for listening.
• 0:23
  SPEAKER_00: If this is your first time, we invite you to subscribe wherever you listen to the podcast.
• 0:28
  SPEAKER_00: And now here are your host, Seth Early and Chris Featherstone.
• 0:40
  SPEAKER_02: Good morning, good afternoon, good evening, depending upon your time zone.
• 0:44
  SPEAKER_02: Welcome to our podcast.
• 0:45
  SPEAKER_02: I'm Seth Early, and uh normally we have Chris Featherstone with us.
• 0:49
  SPEAKER_02: He was not able to make it today.
• 0:50
  SPEAKER_02: Uh, but I want to uh introduce our guest.
• 0:53
  SPEAKER_02: Our guest is a man after my own heart with deep expertise in knowledge engineering, knowledge graphs, and structured content design.
• 1:01
  SPEAKER_02: He spends his time helping mid-sized organizations get their heads and hands around their content and content management.
• 1:08
  SPEAKER_02: Joining us today from Chile, Seattle.
• 1:10
  SPEAKER_02: Please welcome Andy Fitzgerald.
• 1:13
  SPEAKER_02: Thanks so much.
• 1:14
  SPEAKER_02: Great to have you.
• 1:16
  SPEAKER_02: So, Andy, um, we like to kind of jump in and uh get a little bit of your background.
• 1:21
  SPEAKER_02: And I know your uh your degree is in uh you have a PhD in English and literature.
• 1:28
  SPEAKER_02: And how did you go from that to information architecture and knowledge graphs?
• 1:32
  SPEAKER_02: So maybe you can give us a little bit of the background, the world according to uh Andy.
• 1:37
  SPEAKER_05: Sure thing.
• 1:37
  SPEAKER_05: Yeah, it was the short answer is it was a dog-legged, circuitous path, as I think most people in user experience design and information architecture, or many people anyway, are uh come from diverse backgrounds.
• 1:51
  SPEAKER_05: Uh ironically, that uh path started from graduate school from English.
• 1:57
  SPEAKER_05: I decided early on that while I love the academy and I love teaching, I didn't want to spend my whole career there.
• 2:03
  SPEAKER_05: And the my last job as a grad student uh was one that I had no business having uh as a web, uh what was I uh basically a webmaster for a humanities site on campus.
• 2:19
  SPEAKER_05: And I looked at their site and quickly realized that what they need is not web mastering, but they needed information architecture, uh, and basically learned how to do some of what I do now there on the ground.
• 2:31
  SPEAKER_05: I found a mentor that was uh invaluable uh in making sure that I didn't completely screw everything up.
• 2:37
  SPEAKER_05: Uh, and then transitioned into uh user experience design, really, because at the time, 2010, uh there wasn't, there was much more need for broad user experience design than there was for information architecture.
• 2:50
  SPEAKER_05: And the the really pleasant thing for me and the serendipitous thing is the deeper I got into user experience design and the more I continued to grow and develop as an information architect and start to move more and more into that kind of specialization, the more the things I was doing as an English student in grad school came back around because it's all about making sense of how we make sense of the world with words and with language.
• 3:16
  SPEAKER_05: And there's nothing natural uh in the sense of the world about language.
• 3:21
  SPEAKER_05: It's all a lot of construction, it's all very complex.
• 3:24
  SPEAKER_05: And picking those things apart uh really takes me back to the stuff I was doing in grad school, and now that I'm I'm doing it uh as a job, that uh I I like it as much now as I did then.
• 3:35
  SPEAKER_02: That's great.
• 3:36
  SPEAKER_02: And uh and so you really became it was really you were thrown into this, but you found found this opportunity, and then you found that uh looking at this site that you you realize that geez, this needs a little bit different uh way of organizing uh the information.
• 3:52
  SPEAKER_02: And tell tell us more about how that kind of how what your insights were, what was your observation at that time?
• 3:58
  SPEAKER_05: And yeah, it uh was actually there, there was a little bit more of a of a plan than than just discovery.
• 4:06
  SPEAKER_05: I knew going into this job it was uh a human the Simpson Center for the Humanities at the University of Washington, and they had a web editor, was what the role was.
• 4:17
  SPEAKER_05: It was basically someone to update their site, and their site had been around for a good 10 years and it was cobbled together with all sorts of all the bubblegum and straining that you would expect of a site at a humanities center, and they did a good job of keeping it up to date, but it was a very brittle structure.
• 4:34
  SPEAKER_05: Uh, and I uh I expected that uh applying for the job, and actually the the job I had really no business getting was that web editor job because I was an English grad, so Lin uh yeah, uh language and literature.
• 4:46
  SPEAKER_05: Um, but I learned enough of that to be confident in what they needed me to do.
• 4:51
  SPEAKER_05: But I knew going into it that if I could, because I had heard about and learned about information architecture um earlier on and realized, wow, this is about how we make sense of stuff online with words.
• 5:05
  SPEAKER_05: This is what I want to be doing.
• 5:06
  SPEAKER_05: So I knew going into it that if I could, I would like to turn that into a pivot point towards information architecture.
• 5:14
  SPEAKER_05: And I spent it was a year-long appointment.
• 5:16
  SPEAKER_05: So I, you know, spent the first month or so learning the job and making sure that I could deliver on what they hired me to do.
• 5:24
  SPEAKER_05: And then I put together, uh audaciously put together a project proposal and went to the associate director who was my boss and said, Hey, this is what I think we actually need to be doing.
• 5:34
  SPEAKER_05: And I explained this to her and I had a proposal that you know was decent.
• 5:39
  SPEAKER_05: And uh, and I could kind of see the the wheels turning in her her head.
• 5:43
  SPEAKER_05: Uh, and and I imagined that it was something along the lines of wow, this guy wants to do a lot more than we hired him for, and he's not asking for more money.
• 5:51
  SPEAKER_05: Let's let him do it and see what happens.
• 5:56
  SPEAKER_05: Yet, yet.
• 5:57
  SPEAKER_05: Well, and it was the frankly, the learning opportunity was uh it was really the the thing for me that was the pivot into what I wanted to do.
• 6:06
  SPEAKER_02: And so when you saw that, like what was it?
• 6:08
  SPEAKER_02: What were the wheels turning to say, um, you know, I need to look at this differently beyond just you know editing content, I need to from a structural perspective.
• 6:17
  SPEAKER_02: Like what was what would what catalyzed that?
• 6:20
  SPEAKER_05: Yeah, in this case, a lot of it was first of all, the technical infrastructure, that it was a lot of static pages that were updated on the server, lots of stuff in terminal editors, uh, and and then things pulled in from various different blogging platforms that have been set up ad hoc over the years.
• 6:46
  SPEAKER_05: So there was a technical side to it.
• 6:48
  SPEAKER_05: And and then there was the experience side to it.
• 6:51
  SPEAKER_05: But you were interested in the technical side, even with your background of literature and and English, you still have a in- Yeah, I did have well, and and that interest came out of uh just starting to build web pages for my uh undergraduate classes.
• 7:07
  SPEAKER_05: So I was teaching undergraduate uh language and literature, and I built um really rudimentary web pages, but that got me uh it sort of like got me into the the structure of how the web works and just an LM CSS.
• 7:21
  SPEAKER_05: Um but then the other the other piece of the humanity side was that the the menus weren't super easy to navigate, stuff was out of balance.
• 7:28
  SPEAKER_05: So there was a a user experience side to it too that yeah, yeah, exactly.
• 7:35
  SPEAKER_02: And then it's how did you start uh getting into an area of knowledge graphs?
• 7:42
  SPEAKER_02: Because that's one of the areas that we talked about, knowledge graphs and content.
• 7:45
  SPEAKER_02: So how did that kind of emerge for you?
• 7:47
  SPEAKER_05: Yeah, that's been in the last few years and well, maybe the last five, I think.
• 7:54
  SPEAKER_05: Uh, and it's been thinking about content as data versus content as pages.
• 8:02
  SPEAKER_05: And some of that has come out of learning a little bit more about schema.org when that emerged as a way to basically use algorithms to help people parse and find information on the web.
• 8:17
  SPEAKER_05: And uh as we continue to get more and more data and more and more pages and more and more content, it's increasingly necessary to have help to find things.
• 8:27
  SPEAKER_05: That you have a question about uh a cert on a certain topic, and like nobody's gonna sift through just pages and pages of stuff.
• 8:35
  SPEAKER_05: We need the robots to help us out, uh, and that we they can do that better when we have told them in a language that they can understand what the content's really about.
• 8:47
  SPEAKER_05: So schema.org was a first look into that.
• 8:50
  SPEAKER_05: And then from schema.org, it was a uh slippery slope right down the rabbit hole into RDF and thinking about knowledge graphs and thinking about ontologies and how those things fit together.
• 9:02
  SPEAKER_02: So maybe you could talk uh for folks who may be less uh familiar with schema.org, uh, the role that plays, and then talk about how you know that fits in with things like artificial intelligence and you know the the uh the ability to make sense of information at scale.
• 9:19
  SPEAKER_02: When you start talking about yeah, yeah.
• 9:23
  SPEAKER_05: So schema.org is a controlled vocabulary for things on the web.
• 9:29
  SPEAKER_05: And it's rooted in e-commerce.
• 9:32
  SPEAKER_05: Uh, it was originally uh its own entity and it was acquired by Google, and and Google is now, I believe, if I'm not getting this wrong, um, sort of supporting its continued development.
• 9:42
  SPEAKER_05: But a lot of it's around e-commerce.
• 9:44
  SPEAKER_05: So you have I sell shoes online, uh, and my shoes range in sizes from six to ten, and they range in prices from eight dollars uh to thirty-five dollars, say.
• 9:55
  SPEAKER_05: Uh, and all that stuff is things that you can put on a web page, and people understand perfectly well.
• 9:58
  SPEAKER_05: But a robot looking at that stuff just sees a bunch of numbers and text.
• 10:02
  SPEAKER_05: It doesn't have any context for what a shoe is or what a price is or what a size is.
• 10:06
  SPEAKER_05: And what schema.org does is it puts that information in context.
• 10:12
  SPEAKER_05: So you can say, for instance, that uh a shoe is an article of clothing that comes in sizes and uh and relate those things together.
• 10:22
  SPEAKER_05: So, for example, uh an algorithm would know if uh you have a product and it's a shoe, uh, if you have uh something that's marked up as a size related to it, uh schema.org is what allows the algorithm to put those things together so that actually a probably a better example is uh, and this is uh one that I've used in a couple articles and talks, is if I do a voice search on Google for uh gluten-free delivery near me, it comes back with a list of restaurants near my area that deliver food that is gluten-free.
• 10:58
  SPEAKER_05: I haven't said food, I haven't said food delivery, but those terms, gluten-free delivery near me, the what something, and I don't know that they're exactly using schema.org for this, but what something like schema.org allows an algorithm behind the scenes to do is to say, ah, if he's talking about gluten-free and delivery, he's probably talking about a restaurant and it can make some assumptions.
• 11:23
  SPEAKER_05: Right, right.
• 11:24
  SPEAKER_02: Interesting.
• 11:24
  SPEAKER_02: And um, and so when you start thinking about um how knowledge graphs are related to content.
• 11:32
  SPEAKER_02: And so schema.
• 11:34
  SPEAKER_02: is really a structure, it's a sense of records that says here's a way of organizing the information on pages so that uh that that those pages can be interpreted through automated means, right?
• 11:47
  SPEAKER_02: Google can pull that back.
• 11:49
  SPEAKER_02: You can build things like um uh the uh knowledge cards, right?
• 11:54
  SPEAKER_02: Based on the dot org, right?
• 11:56
  SPEAKER_02: So tell talk a little bit about a knowledge card, talk a little about you know how that that structure can be can be leveraged, and then let's put it in the context of you know of how how AI needs to leverage those types of structures.
• 12:10
  SPEAKER_02: Because again, people don't necessarily understand.
• 12:12
  SPEAKER_02: They say things like, well, you know, make it like Google, right?
• 12:16
  SPEAKER_02: Um search internally to an organization.
• 12:19
  SPEAKER_02: And and I always like to say, well, if you kind of think about what goes behind making things work for Google, you would you would look at this a little differently.
• 12:28
  SPEAKER_02: So maybe you could riff on that a little bit.
• 12:31
  SPEAKER_05: Yeah, yeah.
• 12:31
  SPEAKER_05: So the knowledge cards are uh a great example.
• 12:34
  SPEAKER_05: A lot of them come out of Wikipedia, I believe, and Wikidata, uh, and the structured content that comes out of those.
• 12:42
  SPEAKER_05: And you can see that in the link beneath them.
• 12:44
  SPEAKER_05: There's a great quote, it's actually the thesis of the book, Thinking Uh Fast and Thinking Slow by Daniel Kahneman.
• 12:51
  SPEAKER_05: Um, that he talks about what you what you um what you see is all there is.
• 12:57
  SPEAKER_05: And we do this uh and as as humans looking at the world, if we if we don't see an option, we assume it's not there.
• 13:05
  SPEAKER_05: Uh and it's certainly the case for algorithms as well.
• 13:08
  SPEAKER_05: Uh and for something like Google, who's creating a knowledge card, uh, if it can't look at my website and see that I'm an information architect and see that I've been practicing for I don't know, 12 or 15 years or whatever it is, that information is not there.
• 13:22
  SPEAKER_05: I may have printed it on a page, but what I mark it up with something like schema.org, and it's really just it's like it's tagging bits of content and say, hey, this is experience.
• 13:35
  SPEAKER_05: Uh and so what schema.org is, in the same way that I can look at uh someone's LinkedIn profile and see that they have information architecture 12 years.
• 13:45
  SPEAKER_05: I know because of the convention of associating lengths of time, 10 years at a role, with a title, uh, that those two things are related.
• 13:54
  SPEAKER_05: But if I didn't know that, I don't have the context for that.
• 13:57
  SPEAKER_05: There's nothing natural about those things being related.
• 14:00
  SPEAKER_05: So schema.org provides that context.
• 14:02
  unknown: Right.
• 14:02
  SPEAKER_05: And in the AI sense, that's then what allows our machines to be smart.
• 14:07
  SPEAKER_05: We're giving them the context that that we have as humans that are used to working with pages, but they don't.
• 14:15
  SPEAKER_02: Right.
• 14:15
  SPEAKER_02: So, so so talk about um uh, you know, when you start again going back to this analogy where people say, make it like Google.
• 14:23
  SPEAKER_02: Like if I'm in a corporate environment and um I can't find anything, it's like, darn, why can't it just be like Google?
• 14:31
  SPEAKER_02: What do what do we need to help executives understand about that when they say that?
• 14:36
  SPEAKER_02: And and you know, again, when you think about what organizations are putting into optimizing pages so that it is like Google, yeah, yeah.
• 14:46
  SPEAKER_02: Energy in into our internal pages.
• 14:48
  SPEAKER_02: So talk about that.
• 14:51
  SPEAKER_05: Yeah, I I think that's a great question and a great way to frame that problem.
• 14:57
  SPEAKER_05: And and I think that making it like Google means understanding a little bit about what Google does.
• 15:04
  SPEAKER_05: Uh, and and I don't, I'm I'm quickly getting out of the things I actually know, so I'm surmising, but we do know that they crawl the web and they have massive servers that crawl, you know, your web pages and everything that's connected to them every day.
• 15:18
  SPEAKER_05: And the way that they make sense of those web pages is through the semantic structure of HTML.
• 15:24
  SPEAKER_05: So they look at the H1s, they look at the title tags, they look at and often ignore the meta descriptions because we're terrible about making those truthful.
• 15:31
  SPEAKER_05: Uh, and there are all sorts of other signals that they pull in.
• 15:35
  SPEAKER_05: And though that would just be the tip of the iceberg of those signals.
• 15:38
  SPEAKER_05: And they look at schema.org and assemble all of that in a database.
• 15:43
  SPEAKER_05: And and all of that comes back to making information legible to the algorithms.
• 15:52
  SPEAKER_05: And I think with with organizations, they're just using semantic markup, just making sure that your your headings on your web pages, if you want them to be found, are using h1 tags instead of using spans that are large text uh and uh boldface, because that looks like a heading to us, but an algorithm looks at that and says a span, spans nothing.
• 16:14
  SPEAKER_05: Move right on.
• 16:15
  SPEAKER_05: Right.
• 16:16
  SPEAKER_05: So that but making that that legible, I think, is a first step in that.
• 16:22
  SPEAKER_05: And um, well, there was a second point that I forgot what it'll come back up.
• 16:27
  SPEAKER_02: Yeah, so but I think that the the bigger point is that you know organizations spend uh you know hundreds of millions of dollars what the size of the SEO uh uh the search engine optimization marketplace is, but but organizations spend a lot of time optimizing that content so that it can be serviced, so that it is on Google.
• 16:50
  SPEAKER_02: And if we did the same, put a fraction of the same energy into optimizing our internal content in our internet, even paying attention to it to say, let's let's curate it, let's make sure that it that it's uh on topic, let's make sure that it's well written, let's make sure it contains the signals that are necessary for our search algorithms to be able to surface it, but it would be like Google.
• 17:13
  SPEAKER_05: Yeah, yeah.
• 17:14
  SPEAKER_05: Well, it so uh you and I were both recently at the taxonomy boot camp, uh where we've had some you know a number of similarly nerdy conversations.
• 17:24
  SPEAKER_05: Uh and there there are a couple things that that uh come to mind uh for me, and and one was the statement uh that came up in several talks is uh you you can't automatically fix bad data.
• 17:38
  SPEAKER_05: Right.
• 17:39
  SPEAKER_05: If the information isn't there, uh you need to provide it.
• 17:42
  SPEAKER_05: And usually it takes a person to provide it.
• 17:44
  SPEAKER_05: Another one was from Helen Lipple, who's a taxonomist and who was on the stump the taxonomist panel on the last day.
• 17:51
  SPEAKER_05: And someone in the audience, not me, asked, How do you keep people from putting everything in the miscellaneous category?
• 17:58
  SPEAKER_05: Or how do you how do you keep a miscellaneous category from coming up?
• 18:02
  SPEAKER_02: General, other miscellaneous category.
• 18:05
  SPEAKER_02: Exactly.
• 18:07
  SPEAKER_05: Yeah, exactly.
• 18:08
  SPEAKER_05: Otherwise, otherwise uncategorized, yeah.
• 18:10
  SPEAKER_05: And response I really loved.
• 18:13
  SPEAKER_05: Uh, a couple, a couple other people took a stab at it.
• 18:15
  SPEAKER_05: None of them were wrong, but her response was if you have a lot of things in the miscellaneous category, your domain model is probably wrong.
• 18:22
  SPEAKER_05: Right, right.
• 18:23
  SPEAKER_05: That you probably haven't created a structure uh to allow people to put things in those categories.
• 18:29
  SPEAKER_05: And and that makes me think, as we're talking about this, about work that I've done for intranets for city government.
• 18:37
  SPEAKER_05: And uh, you know, they'll call in a consulting agency to fix their internet because people can't find anything and they spend, they waste hours and hours and end up just standing up and shouting over the cubicle, hey, where do I find this thing?
• 18:51
  SPEAKER_05: Uh and a lot of, and this is the the bane of intranets everywhere.
• 18:54
  SPEAKER_05: And I think a lot of it is because people are they're allowed to self-organize and they become a thicket of unfindable information beset by thorns on every side, and nobody wants to go in there.
• 19:07
  SPEAKER_05: And some of that is that there isn't a taxonomy, uh, and there isn't a model for organizing things.
• 19:13
  SPEAKER_05: So there's not a findable place to put things.
• 19:16
  SPEAKER_05: And I think when uh thinking about organization and taxonomy in particular, of which schema.org is one, gets left until later or uh not thought of at all because uh Oval applying natural language processing to it, uh, which can fill some gaps, but it's a really hard and expensive way to do it.
• 19:38
  SPEAKER_02: You're making up for your past sins in core configurations.
• 19:44
  SPEAKER_02: You're trying to take a sledgehammer and you know, and kill what would have been a fly uh when you just tried to kill that fly earlier on, right?
• 19:54
  SPEAKER_02: Now you're saying we have all this with this chaos and all this mess.
• 19:58
  SPEAKER_02: So I'm gonna try to apply uh machine learning algorithms and I'm gonna try to apply uh natural language processing, and I'm gonna do a lot of work, you know, source structures and all this.
• 20:08
  SPEAKER_02: And really at the end of the day, it's because you didn't fix this stuff in the first place, right?
• 20:13
  SPEAKER_02: You didn't have a structure and you didn't have the right domain model.
• 20:16
  SPEAKER_05: Yeah, yeah.
• 20:18
  SPEAKER_05: Yeah, and there's uh, I mean, we've seen a couple examples.
• 20:21
  SPEAKER_05: The um recently, the the KFC uh crispy chicken Kristallnacht commemoration is one particular.
• 20:31
  SPEAKER_05: Oh, oh yeah.
• 20:32
  SPEAKER_05: So this was a couple weeks ago.
• 20:34
  SPEAKER_05: Um the KFC ran a promotion in Germany commemorating Kristallnacht, which was the uh this horrible event of violence against Jews uh in the mid-century, uh, by suggesting people buy a cheesy crispy chicken sandwich.
• 20:52
  SPEAKER_05: Oh my goodness.
• 20:53
  SPEAKER_05: Oh yeah, indeed.
• 20:55
  SPEAKER_05: Oh my goodness.
• 20:56
  SPEAKER_05: And yeah, yeah.
• 20:58
  SPEAKER_05: And of course, uh the internet went crazy over this.
• 21:01
  SPEAKER_05: And KFC immediately issued an apology.
• 21:06
  SPEAKER_05: And uh you know, you there's a an art, a brief article on the BBC, uh BBC News about this, just saying that it, you know, it was a mistake, it was a bot, we had something wired up and it went out, it should never have gone out.
• 21:19
  SPEAKER_05: We deep, of course, wildly embarrassing and horrible press for them.
• 21:24
  SPEAKER_05: But that's yeah, that's the automation.
• 21:27
  SPEAKER_05: And I, you know, as we talk about what do you do when you're so if you've got hundreds of thousands of documents, some some automation for for tagging for auto-tagging is probably a good idea, but you but there's always a human element, uh what the bull party guys call the human in the loop, uh that to verify that so that you don't do something like recommend that people buy crispy chicken sandwiches for some cataclysmic event.
• 21:55
  SPEAKER_02: That's pretty that is that is pretty horrifying.
• 21:58
  SPEAKER_02: That is really horrifying.
• 22:00
  SPEAKER_02: crazy.
• 22:01
  SPEAKER_02: So you know, when you think about, you know, talk a little bit about knowledge graphs and for the audience, define knowledge graph uh in your in your words.
• 22:09
  SPEAKER_02: And then how does that help us with content?
• 22:12
  SPEAKER_02: And then how does that help us with AI?
• 22:16
  SPEAKER_05: Yeah.
• 22:17
  SPEAKER_05: So a knowledge graph is I think the shortest way I can put it is con data in context.
• 22:27
  SPEAKER_05: So if I have a set of data, let's go back to my imaginary shoe store, right?
• 22:33
  SPEAKER_05: I've got data or let's let's talk about REI.
• 22:35
  SPEAKER_05: So REI has got a lot of data about a lot of articles of clothing.
• 22:40
  SPEAKER_05: And it that data all on its own doesn't have the kind of context that even something like schema.org, which is a really good place to start, can provide it.
• 22:50
  SPEAKER_05: So it's they're graph based uh in in a sense for boy this is a a big topic to to open up.
• 23:00
  SPEAKER_05: So your data is in let's say a relational database what are often called or referred to as closed world systems, where is if something isn't in the database, it comes back as a false value.
• 23:13
  SPEAKER_05: In um web ontology and in graph databases, there is this idea of the open world assumption.
• 23:21
  SPEAKER_05: And the idea of that is that absence of evidence is not evidence of absence.
• 23:26
  SPEAKER_02: Right.
• 23:28
  SPEAKER_05: So uh what a knowledge graph is data information in context.
• 23:35
  SPEAKER_05: So what they what they usually consist of is a data set like your inventory and then a taxonomy or ontology which says um hey if it's uh a shoe and it is uh I'm sorry that maybe the shoes aren't the best example um oh well here we go if it's a shoe it has to have a size and it has to have a color.
• 23:58
  SPEAKER_05: So there are a couple things you can do with that.
• 24:00
  SPEAKER_05: One is you can use that to validate the information going in and say hey you can't enter a new product and call it a shoe and not give a size or color.
• 24:09
  SPEAKER_05: That's not possible.
• 24:10
  SPEAKER_05: It doesn't exist.
• 24:11
  SPEAKER_05: And then you can also um use that on the when you're publishing or you're expressing that information and you're publishing it to a web page you can say hey I've got a shoe I know it's got a size and I know it's got a color the size is marked up with the schema colon size property and the color schema colon color.
• 24:33
  SPEAKER_05: And then you can start to do things once you have that data in place you can use the context to say show me all the brown shoes right or show me all the shoes that are in inventory um yeah uh in in you know in size 10 or something um there's the the the other example there's a client that I'm working with with um right now that is that I'm helping build a structured content solution and and I think here we'll start to to to broaden this out a little bit uh and they are uh point of care microlearning for healthcare providers so specific information say for nurses specific to their unit uh and uh this organization creates a lot of custom content and by being granular about how we describe the elements in that content and then relating them together now we can look at the data set as a whole and say okay we have a new client who is a NICU level three um I want to query all of our hundred other hundreds of other clients who are NICUs and level three and see what what pieces of content do all of them have?
• 25:46
  SPEAKER_05: What pieces of content do 90% of them have what pieces of content 80% of them have now I can go back to my new client and I can say hey we're really confident you're gonna want this information we're somewhat confident you're gonna want this information you may also want this and it's a way to basically mine our content set as data.
• 26:06
  SPEAKER_05: And the the graph part of it is in that open world assumption that I mentioned a moment ago so um the idea that there can always be new information means that we then build in the possibility of then layering in traffic data.
• 26:21
  SPEAKER_05: So if for instance we can say show me not only the resources that everybody has but show me the of the units that are the most like this new unit coming in what are the resources that have the highest traffic and if we were doing this with relational databases you could do some joins and you could probably make it work but it's work.
• 26:40
  SPEAKER_05: If you're doing this with a knowledge graph if you uh and this is part of the the beauty of working with open standards like scope or if you then say uh hey the this entity X in my analytics data set is same as is the same thing as entity Q over here in my records database and then using a graph query Sparkle um or or other kinds of queries um you can query that entire combined data set so it's a it's a way of uh using the data in multiple ways um by making and and really uh leveraging algorithms uh to help us to allow us to use the data in multiple ways by describing them not as pages um but actually data elements that have relationships right right right so you have the ability to you know query this uh the this information in ways that you did not anticipate and that you didn't in advance and also to navigate from you know uh from from node to node or I like to think of a knowledge graph as as uh you know or the ontology is the scaffolding of information and then when you we take that scaffolding and we marry it up with the data that's when we have a knowledge graph and then that knowledge graph the ability to uh to uh find new relationships and to be able to navigate that that knowledge structure in ways that we did not anticipate or that we did not plan for right like yeah node to node and looking for the things that are in common um you know you you had uh you'd mention the medical application but again you could going back to your shoe example you know you could say well what all what other red things do I have right looking at the attribute the attribute can be the key and to other types of of entities right and so we have lots of kind of navigate that so playing the same degrees of Kevin Bacon on IMDb all of your products all of your knowledge all of your your resources all of your entities yeah right right now I had a go ahead sorry love that definition and I've also been not avoiding the the AI question on purpose uh I I think there did before you go to the next question to to start to answer some of that I think uh the the implication for AI is providing context there is uh I another set of thinking that has pretty profoundly influenced how our practice information architecture is cognitive linguistics and I think um yeah so um the idea that our understanding of the world is based in metaphor uh we understand things we understand new things based on old things that we understand and uh there are some thinkers um uh like George Lakoff uh Mark Johnson who who actually walk that all the way back to embodiment uh and that I we understand the world because we have bodies so they talk George Lakoff in Women of Fire and Dangerous Things talks about what he calls kinesthetic image schemas as being the first thing that we understand to understand other things.
• 29:57
  SPEAKER_05: And the kid kinesthetic image schema is something like um I have a body so I understand inside and outside because I know what's in me and I know what's out of me.
• 30:07
  SPEAKER_05: I understand um part and whole uh because my I am I am a whole being and if part of my whole being is made part it's upsetting and I have a really I don't like when that happens.
• 30:18
  SPEAKER_05: I don't I don't like when that happens either it's my least favorite thing.
• 30:21
  SPEAKER_05: Yeah it's believed yeah good good I'm glad to hear it so those uh of course they seem very very basic but like often Johnson argue that that's how all of our understanding is built up that everything comes from those beginnings and if you spend any time around babies I don't very often um but like looking at them explore and we we we forget about you know the first 10 years of our lives that we're figuring out how to understand things and how absolutely nothing we understand at the very beginning.
• 30:55
  SPEAKER_05: So then that starts to put these image schemas and the way that metaphor actually builds our entire way of thinking in the world.
• 31:04
  SPEAKER_05: And that is something that to be maybe a little uh philosophical about it, an AI until it has a body it's missing that foundation it's missing that context.
• 31:16
  SPEAKER_05: And when we can provide it and we can um sort of like shortcut some of that learning process by providing things like schema.org or by providing ontologies we can provide the context that we forget we ever generated you know it's interesting I was talking to one of our um one of our uh podcast uh participants um uh who talked about the idea of he believes that that we will at some point have uh you know have solved the problem of general artificial general intelligence you know and artificial general intelligence is when you know when we have machines that are at the level of parity with with humans and yeah I I find that a little bit hard to believe a lot there you know because we know how stupid these things are and how you have to teach them everything.
• 32:13
  SPEAKER_05: And yeah your perspective on you know cognitive uh uh computing cognitive applications cognitive assistance and content right because to me it's all about content and it's all about context right people are trying to build these these uh applications without necessarily leveraging things like knowledge graphs without leveraging schemas and architecture the knowledge architecture there there you still see a lot of very very basic uh instantiations of these things that are kind of somewhat hard coded right there pairs there's they're not necessarily maybe using knowledge bases but not necessarily using knowledge architecture or information what are you seeing in that space yeah I certainly similar things there was the uh with the Microsoft chatbot from a few years ago back in the before times I forget what it was Tay or um yeah yeah that over a period of like three or four days on the internet became like a you know Nazi like it's just completely hate yeah yeah yeah yeah uh so we there's his great great grandson right what yeah went off the rails and became a skinhead yeah yeah yeah so we we see that and I think uh I think some of it is I mean I think it's a result of wildly underestimating how complex thinking is and how complex linguistic processes are because they're so easy to us of course we feel like oh I'm I'm a I'm good at coding I can code that um some of the six some of the more um successful instances I think are when those kinds of applications have a better awareness or they communicate a better awareness of their limitations.
• 34:07
  SPEAKER_05: So um Siri for instance is uh you know was launched and still is meant to be very personal and very human like um but you could I I don't use that tool very often because it doesn't do a lot of things that I can't just do on my own easily.
• 34:22
  SPEAKER_05: It's not useful for me.
• 34:26
  SPEAKER_05: Does turn on the lights for me when I'm far away from the light switch there's there's that uh well you know anyway yeah yeah there are there was uh I think it was one of the early Google products um and as we talked about it'll it'll come back to mind um was much less human like um and so we expected much less of it it could do mostly the same things that Siri could do um but when it failed it was like well of course it doesn't do that right i'm I'm working with a more rudimentary thing whereas Siri and the I think the the attempts to be cheeky or to be clever or to mimic human characteristics to get people comfortable with it.
• 35:09
  SPEAKER_05: But then when you when you ask it to do something a little more complex like uh you know try the lights off in a half hour uh and it does something completely unexpected because you've added just just enough complexity that it it can't handle it uh it's being stupid because it promised you it was smarter.
• 35:27
  SPEAKER_02: It's funny I was asking a question I was trying to get to a live agent when I had a problem with um Delta Air and I ended up and I figured it's going to pass it on to a human at some point.
• 35:41
  SPEAKER_02: So I asked this long question and the bot responded with I'm sorry that question is too long for me can you simplify it and I said no this is a complex situation right I had this experience and I need a human yeah and then I got a human and the human told me to go back to where I started you know with the whole and that bot again and that bot and it just cycled me through this and it was oh man it's really so horrendous.
• 36:10
  SPEAKER_02: And you still see these really really poor you know degradations of these applications you know where you you're missing an apostrophe right and it doesn't you're missing a dash and it doesn't it's like guys you're not even using you're not even using machine learning you're not even using NLP you're you're you're completely off the mark here because you're not using that's using regex yeah right it's like it's it's ridiculous yeah yeah yeah and so um so uh so when you start thinking about um you know ai it to me it is a content problem at the end of the day you know yeah I I had a note here uh talking about boutique knowledge graph but it but I don't recall uh the context of that discussion yeah so uh oh and by the way I should say that um we're a little past our midpoint but we are with Andy Fitzgerald who is uh based in Seattle he's an information architect and user experience expert we'll be right back we'd like to thank Marketing Artificial Intelligence Institute for their support of this podcast.
• 37:19
  SPEAKER_02: Marketing AI Institute is a media event and online education business that makes AI approachable actionable and accessible to marketing leaders around the world they run the world's leading marketing AI conference at MACOM which brings together hundreds of marketing and business leaders to better understand how to pilot and adopt AI today's podcast is brought to you by Simpler Media Group, owner of the Customer Experience Decoded podcast covering the latest in customer experience and marketing SEX Decoded is a semi-monthly podcast designed to deliver insight from leaders driving the evolution of customer experience with an emphasis on digital touch points, customer journeys, respectful data management and voice of the customer I want to give a big thanks to Early Information Science in support of the podcast.
• 38:05
  SPEAKER_01: Early information science is a professional service firm with proven methodologies designed specifically to address product data, content assets, customer data and corporate knowledge bases.
• 38:14
  SPEAKER_01: Their information architecture expertise has allowed clients from all over the world to efficiently organize, access and utilize their data, ultimately resulting in better business outcomes.
• 38:24
  SPEAKER_01: They deliver customer driven solutions that scale and adapt to the needs of a business as it evolves.
• 38:30
  SPEAKER_01: Founded in 1994, early information science has proudly delivered result-driven solutions for more than a hundred brands across the top industries.
• 38:38
  SPEAKER_01: You can get more information about Early Information Science at early.com early dot com.
• 38:46
  SPEAKER_01: Working with customers in AI, the number one question that I get usually is where do I start?
• 38:52
  SPEAKER_01: What data do I need?
• 38:53
  SPEAKER_01: How do I gather it?
• 38:54
  SPEAKER_01: And it's great questions.
• 38:56
  SPEAKER_01: Well if you want to catch the wave of AI, not be crushed by it, you really do need to read a book written by Seth early called The AI Powered Enterprise.
• 39:04
  SPEAKER_01: It gives you the ability to look and understand how to harness the power of ontologies to make your business smarter, faster and more profitable.
• 39:12
  SPEAKER_01: Seth goes into the intricate details of how to build an ontology, develop and employ an ontology as a secret weapon for the basis for how to actually utilize AI as a powerful competitive advantage.
• 39:24
  SPEAKER_01: This book is really the first one to combine sophisticated explanation of how AI works with a practical approach of applying AI to the problems of the business.
• 39:33
  SPEAKER_01: And it's applied to you know various use cases from customer experiences to business operations to product development.
• 39:39
  SPEAKER_01: It's a powerful powerful tool in the arsenal of understanding how to deploy AI to make your business better.
• 39:45
  SPEAKER_01: The AI Powered Enterprise by Seth Early harness the power of ontologies to make your business smarter faster more profitable out on Amazon.com today.
• 39:53
  SPEAKER_01: And now back to the show thank you for that reminder.
• 39:59
  SPEAKER_05: Yeah so boutique knowledge graphs are is um an experiment uh of mine that brings together many of the things that we've been talking about so structured content knowledge modeling ontologies and knowledge graphs at a small scale and you you might be wondering well why at a small scale often knowledge graphs and ontologies get applied to very large data sets.
• 40:28
  SPEAKER_05: So they're I think most of the most of the ontology applications that I've seen examples of are in biotech where you're you're tracking proteins or you're tracking you know uh drug interactions and these very complex huge data sets where you need uh tons of computing power in order to figure things out the the open world assumption and connected content and linked data brings a whole host of advantages to those kinds of applications.
• 40:56
  SPEAKER_05: Many of those advantages like being able to uh combine data sets like we talked about earlier or being able to generate a schema uhorg markup for your content are also benefits that smaller organizations can uh see see value from and so my the the the reason behind this experiment or the the motivation behind it was to say well what would a small scale knowledge graph look like and would it be worth doing or would it just be so much work that you would be you know it would be that hammer to the fly that you're talking about earlier.
• 41:35
  SPEAKER_05: So UX methods comes out of a challenge that I saw in teaching users uxmethods.org so tell us ux methods yeah tell us what that is yeah so uxmethods.org is a website that uh it's a website and knowledge graph that communicates the relationships behind the different methods that ux practitioners use in their craft so the the you can go to the site's live you can go to uxmethods.org and check it out uh the motivation for the site comes out of when I was teaching information architecture to uh students at a UX design certificate course at the School of Visual Concepts here in Seattle and I would teach them about card sorting teach them about tree testing teach them about navigation design and they got all the individual concepts they understood the techniques they knew how to use the tools uh they they could do exercises but the thing that consistently over several courses stumped them was how do I string these things together?
• 42:36
  SPEAKER_05: How do they relate?
• 42:38
  SPEAKER_05: And so what UX methods does the the the premise behind it is it provides an overview of the methods of user experience design of for which there's no shortage on the internet right that's that's not a problem it's solving but what the knowledge graph does is it connects the different methods.
• 42:58
  SPEAKER_05: So what the there's a very basic ontology behind it and the ontology is all about outcomes and inputs.
• 43:06
  SPEAKER_05: So card sorting for example uh creates and I'm just gonna go to the side actually and I can um tell you exactly what we're talking about.
• 43:19
  SPEAKER_05: So if we have um card sorting as an output or as an outcome creates insight into mental models and user preferences.
• 43:31
  SPEAKER_05: There are things other methods like category design or taxonomy design or wireframing that use mental models and user preferences as inputs.
• 43:42
  SPEAKER_05: Card sorting also has inputs uh that are generated by things like user interviews, task analysis and stakeholder interviewing.
• 43:50
  SPEAKER_05: So the idea as an experienced UX practitioner, I know these things as someone who is coming into it it's Harder to put together.
• 44:01
  SPEAKER_05: So what the what UX methods does is it captures each of these methods, provides an explanation, provides a set of steps, but also captures the data around what the inputs and the outputs are.
• 44:12
  SPEAKER_05: And then what the knowledge graph does is for each method.
• 44:18
  SPEAKER_05: So let's take continue with card sorting, it mines the data, it calculates which other methods provide the greatest number of inputs to card sorting, which provide the greatest number of outputs that are used by other kinds of methods.
• 44:36
  SPEAKER_05: So that's what the knowledge graph does.
• 44:40
  SPEAKER_05: And you could do this with um front-end templates as well.
• 44:46
  SPEAKER_05: But as you continue to add content, the calculations get more complex.
• 44:50
  SPEAKER_05: And with by bringing it out into a knowledge graph, if I want to start to weight these algorithms, or if I want to start to bring in, say, Google Analytics data and say use that to rank them by the most visited concepts or um other data that they can start to blend data because of that open world assumption.
• 45:10
  SPEAKER_05: So that's the idea behind it.
• 45:17
  SPEAKER_05: Is there value that you can add um by using a knowledge graph?
• 45:21
  SPEAKER_05: And is setting it up more work than the value you get from it.
• 45:27
  SPEAKER_05: Right, right.
• 45:28
  SPEAKER_02: And so as you build out the content and as you add content, when you have graph relationships, you're not having to kind of reconnect the other pieces, it's building and it's adding to that structure.
• 45:42
  SPEAKER_02: And the inherent relationships are already there, right?
• 45:46
  SPEAKER_02: Yeah.
• 45:47
  SPEAKER_02: So what is your what is your sense so far?
• 45:50
  SPEAKER_02: Is that is the juice worth the squeeze?
• 45:54
  SPEAKER_05: So far, yes, yes.
• 45:56
  SPEAKER_05: And the uh part of what was interesting about doing this is that one of the things that I was uh testing in in creating this was is it worth creating a small model?
• 46:07
  SPEAKER_05: Uh is it cost effective to create a small model?
• 46:10
  SPEAKER_02: Well, yeah, and I'll have I have to tell you, um, I totally agree with the premise because if you start thinking, and and the way I kind of describe this is thinking about a knowledge architecture and a subset of a broader information architecture.
• 46:25
  SPEAKER_02: So if we wanted to build a cognitive assistant, we don't need every element of that enterprise IA, right?
• 46:32
  SPEAKER_02: We need a focused subset, right?
• 46:35
  SPEAKER_02: We need to really look at what are those use cases and what are those scenarios and what's that content that needs to be uh um uh extracted or retrieved based on you don't need a very big complex knowledge rack, right?
• 46:50
  SPEAKER_02: You don't need a complex ontology, you need a streamlined ontology more information because you're really trying to talk about retrieval in these contexts, right?
• 47:01
  SPEAKER_02: What gets me is when people have these giant ontology and I just added like I I did a project with Joint Task Force North many years ago, JTF North, uh, which is uh one of the um uh commands to that actually uh retrieves the uh uh that that protects the US.
• 47:20
  SPEAKER_02: And one of the interesting things is that we built a fairly focused um ontology, and then someone said, you know, I added in a whole new military uh ontology to this because you know it had all these other relationships.
• 47:34
  SPEAKER_02: And now maybe there was a reason for that at the time, but to me it was it might have overcomplicated things, right?
• 47:40
  SPEAKER_02: Because it now big and very bulky and very complex, whereas you have to look at what the fitness to purpose was.
• 47:46
  SPEAKER_02: Then maybe there were some use cases that made sense to more complex than the initial use cases that we had talked about, right?
• 47:54
  SPEAKER_02: Yeah, kill for those.
• 47:56
  SPEAKER_02: So I believe that a lighter weight ontology, a lighter weight knowledge would be exactly what is needed when you're trying to build these retrieval mechanisms, right?
• 48:06
  SPEAKER_02: So it doesn't have everything.
• 48:08
  SPEAKER_02: And so you get some of these knowledge models that are very, very complex, very, very heavyweight, trying to add more and more, and then you say then you start seeing relationships in there that are like irrelevant to the use case.
• 48:21
  SPEAKER_02: And then how do you how do you uh how do you um uh uh wayfind when you have such a huge, you know, complex structure, right?
• 48:31
  SPEAKER_02: How do you accept that?
• 48:32
  SPEAKER_02: So I think it's I think there's a lot of value to doing it though.
• 48:35
  SPEAKER_05: Yeah, yeah.
• 48:37
  SPEAKER_05: Well, and the um, so there are a couple nuggets of wisdom come to mind.
• 48:42
  SPEAKER_05: Um, one is from Jim Hendler, who uh uh co-authored with Tim Bernersley the uh 2001 uh semantic web paper.
• 48:49
  SPEAKER_05: Uh and his and is a professor who's written a lot about this.
• 48:53
  SPEAKER_05: Um likes to say that a little semantics goes a long way.
• 48:56
  SPEAKER_05: Yeah.
• 48:57
  SPEAKER_05: And the other one is from uh John Gall, who is the author of a funny little book in the 70s called Systematics.
• 49:02
  SPEAKER_05: And he uh quotes uh quotes, he creates what he calls Gaul's Law, or what we now call Gaul's Law, that is uh a complex system that works will invariably uh be found to have emerged from the simple system that works.
• 49:16
  SPEAKER_05: Interesting.
• 49:17
  SPEAKER_05: And that that simple system that works is what you're talking about.
• 49:21
  SPEAKER_05: Yes, yes, and I think there's a place in the open world assumption and in graph databases that turns uh the uh accepted thinking about creating databases on its head.
• 49:35
  SPEAKER_05: And that's you're creating a relational database and you create something small and then you add to it and you add to it and you add to it, by adding more tables and more databases, you get a mess.
• 49:45
  SPEAKER_05: Right, exactly.
• 49:46
  SPEAKER_05: If you're using a graph database and you create something small and then you add to it and you add to it and you add to it, you still just have a graph database.
• 49:53
  SPEAKER_05: Right, that's interesting.
• 49:54
  SPEAKER_05: Yeah, they're built to grow.
• 49:55
  SPEAKER_05: Yeah, yeah.
• 49:56
  SPEAKER_05: That's really interesting.
• 49:57
  SPEAKER_05: Yeah, yeah.
• 49:58
  SPEAKER_05: The other thing that fascinates me about starting small is A, if you have a small system that works, you've got a system that works.
• 50:06
  SPEAKER_05: And if it provides value, you're providing value.
• 50:09
  SPEAKER_05: Is that the technologies behind uh really the like the powerhouse tools like um pool party and like some of the other uh you know knowledge graph tools that have you know enterprise price tags, they're using RDF, RDFS, Owl, hardly any owl, but scos behind the scenes.
• 50:27
  SPEAKER_05: You can use these open standards, are out and available for anyone to use.
• 50:32
  SPEAKER_05: So you can start small, and UF Methods uses a tiny, tiny bit of OWL, um RDF, and then some SCOS.
• 50:41
  SPEAKER_05: But if I outgrow the it's you it's um using data.world to run inferences right now, which is a great tool.
• 50:47
  SPEAKER_05: Uh it's fun to really fun to model this kind of stuff on.
• 50:51
  SPEAKER_05: Um but if I outgrow that and I say, hey, I need pool party, it's using the same standards that I use now.
• 50:56
  SPEAKER_05: So if I or if I need to move it on to Amazon and Neptune, if as long as it's standards based, it can grow.
• 51:02
  SPEAKER_05: So you're right.
• 51:03
  SPEAKER_05: So it's it's not even that starting small is a budget way to do it.
• 51:07
  SPEAKER_05: I think it's just the smarter way to do it all around.
• 51:10
  SPEAKER_02: It's a very good point because I think you you do want to start with focused use cases, focused scenarios, and a and a focused IA, and not you know, solve all problems and boil the ocean, right?
• 51:21
  SPEAKER_02: It it really is about some simplicity.
• 51:24
  SPEAKER_02: And I like to say that you know, you know, the simple simplicity is hidden complexity, right?
• 51:31
  SPEAKER_02: But don't start with too much complexity, right?
• 51:35
  SPEAKER_02: Too much complexity.
• 51:36
  SPEAKER_02: Okay, a lot of times, you know, I like to say this is this this is very simple, but there's a lot of complexity behind it.
• 51:42
  SPEAKER_02: My my my iPhone.
• 51:45
  SPEAKER_02: But uh, but we hide that from the user when we're trying to do this.
• 51:49
  SPEAKER_02: But I think I think you know the point here is start as simply as possible, right?
• 51:54
  SPEAKER_02: And and uh you know, accommodate a broader range of uh use cases, you kind of incrementally grow the capability and then build on that simpler knowledge to add yeah.
• 52:07
  SPEAKER_05: Yeah, yeah, I like to think of that um simplicity and complexity thing in terms of clarity.
• 52:13
  SPEAKER_05: And I get this from Abby Covert, who's a college architect, um who uh the that the iPhone is very complex uh and but it's clear.
• 52:27
  SPEAKER_05: And if you take something that's complex and uh and this paraphrasing, I think, some of Abby's thought, you take something that's complex and you simplify it, you st you start to remove simplifying is removing functionality, is removing the ability to do something.
• 52:42
  SPEAKER_05: If you clarify, as Apple has done, a beautiful job of doing for uh consumer electronics, you can then have that complexity, uh, but without uh burdening people with complicated trying to reset the clock on your VCR kind of instructions.
• 53:00
  SPEAKER_02: Yeah, that's so interesting.
• 53:02
  SPEAKER_05: For the kids in the audience, a VCR is like a tape recorder for movies, and a tape recorder is never mind.
• 53:09
  SPEAKER_02: You used to actually have to get up and change the channel, you know.
• 53:13
  SPEAKER_05: Oh, oh, how do we Andy?
• 53:17
  SPEAKER_02: It's been a pleasure.
• 53:18
  SPEAKER_02: It I always enjoy our conversations.
• 53:20
  SPEAKER_02: This has been a lot of any uh final thoughts for folks kind of going down these paths and uh trying to um build out their uh their information uh structures and their uh and and building out knowledge bases and content systems.
• 53:34
  SPEAKER_02: Anything, any final uh words of wisdom?
• 53:38
  SPEAKER_05: I will uh yes.
• 53:40
  SPEAKER_05: Uh and this this isn't directly a plug for for people like you and me that do this kind of work.
• 53:46
  SPEAKER_05: Uh, but one of the I'll pass on some wisdom that that I've heard of from people who have tried to put together knowledge graphs or have tried to put together these complex systems, um, or even even taxonomies.
• 53:58
  SPEAKER_05: Um, hire people that know what they're talking about.
• 54:00
  SPEAKER_05: And I don't I don't say that glibly, but a lot of I've seen a lot of taxonomy projects where there was someone in house who was interested in doing it, yeah, and they say, let's give it a try.
• 54:10
  SPEAKER_05: Um, and and then you know ended up spinning and frustrating a lot of people, and then bringing in someone that had some expertise uh that can that can save a lot of headache.
• 54:21
  SPEAKER_02: I think I would have add a corollary to that.
• 54:24
  SPEAKER_02: Yes, bring someone in who knows what they're talking about who can explain it to you in plain language because if you understand it, just move on because I've had many, many people who call themselves ontologists.
• 54:41
  SPEAKER_02: And most of the information architects or taxonomists, they call themselves, and then they speak about things in very abstract terms that are very confusing and very difficult to understand, and it sounds like they're smart, but in some cases it's not because they're smart, it's because they're confusing.
• 55:01
  SPEAKER_02: I'd like to say, I just add to that, knows what they're talking about and can explain it to you in simple.
• 55:09
  SPEAKER_05: In ways that you understand.
• 55:10
  SPEAKER_05: Yes, I completely agree with that.
• 55:11
  SPEAKER_02: Yeah.
• 55:12
  SPEAKER_02: Thank you so much for your time today, Andy.
• 55:14
  SPEAKER_02: For having me.
• 55:16
  SPEAKER_02: So um uh that's it for today, and uh, we'll see uh everyone next time.
• 55:25
  SPEAKER_00: Thanks for listening to the Early AI podcast with your hosts, Seth Early and Chris Beatherstone.
• 55:31
  SPEAKER_00: We hope you enjoyed the program and took away some ideas that you can start to implement in your organization right now.
• 55:37
  SPEAKER_00: If you're stuck and need help, check out www.early.com for information about early information science and testing services.
• 55:46
  SPEAKER_00: And of course, check the show notes for links to any resources we mentioned today.
• 55:50
  SPEAKER_00: That's all for now.
• 55:51
  SPEAKER_00: See you next time.