 All right, I think we're going to get started. And if I don't get too close to that, I think it's the answer. Good afternoon, everyone. Welcome to our session. We're going to be talking about cultural and technical transformations, benefits from Yale's linked data cross-collection discovery platform. I'm Susan Gibbons. I'm Vice Provost for Collections and Scholarly Communications. It's also important to know that I've had about 25 years as a librarian coming into this project. And I'm joined by my colleague, Rob Sanderson, who is Director of Cultural Heritage Metadata Projects at Yale. We're going to split this talk up. I'm going to talk about, well, the success of this project really has two pieces, technology and culture. I'm going to talk about the culture part, and I'll leave Rob to talk and explain the technology. So a little bit of a grounding at Yale. We have three museums and a library. Our Yale University Art Gallery is the oldest university art gallery in the country and one of the largest. We have the Yale Center for British Art, which is the largest collection of British art outside of the UK, brought to us by Paul Mellon. We have the Yale Peabody Museum, which is a natural history museum as well as anthropology museum. And then Yale University Library, which is the second largest library in the US. And so our challenge is how do you take these very, very large institutions with tens of millions of items and a staff collectively of about 800 people and work on a project together? What's important to know is that at Yale, teaching with collections is our bread and butter. We feel that an education at Yale is qualitatively different because of the opportunities that our students have to engage with the collections from across the museums and the libraries and the galleries. And to give you an example or a scale, by the end of this academic year, COVID willing, we'll have had over 1,000 classroom sessions in a special collections or object classroom throughout the Yale campus. So it happens at a very, very large scale. But what doesn't happen is teaching that way across the collections. That teaching happens in the silos. So we have classes in the art gallery, we have classes in the library, but rarely do those classes connect together. Rarely are the collections that are being taught come from across the collections. And that means lost opportunities. And I wanna give you an example of one of those lost opportunities. So we had a class, a medical school class that came to the art gallery to study these lithographs that have watercolor paintings on them that are by the contemporary artist Jocelyn Gardner. There's a long complicated story to these images and I don't have time to explain them all today, but there's three elements to each picture. It is the back of a head of an African slave woman intricately braided. There is a collar around her neck, which you can see is quite torturous. And then there's a plant that is at the bottom. And that's a medicinal plant. And so it is the combination of those three things that Jocelyn Gardner brought together in her images. So the class came to the art gallery in 2018. They looked at these seven pictures, they asked questions of it, and they left. And that was the end of the story. But there was a whole bunch of lost opportunities here. And because Jocelyn Gardner's collection actually has 13 images, there's 13 lithographs. The other six were literally across the street at the Yale Center for British Art. But the art galleries catalog and the Yale Center for British Arts catalog do not reference the collections that are across the street. So only a few of us happen to know that both collections have these images. If you have the opportunity to read Jocelyn Gardner's artist statement, you learn that she was inspired to create these images by reading the diary of Thomas Thistlewood. Thomas Thistlewood was a plantation owner and a slave owner in Jamaica. And these diaries, there's about 62 in number, cover the year 1750 to his death in 1786. Where are the diaries? The diaries are two blocks away in the library. You would find out that they're in the library if you tried and searched the library's catalog. You'd find the images, which you see one on the screen here, if you went into the library's digital library and saw the images. So now we're up to four different catalogs. Now the students came in part to understand the plants that are in the picture because they are having medicinal value. Well, of course, with a Peabody Natural History Museum as you would expect, we have the plant specimens and they're in another museum and they're in another catalog. So to find all of this would meant a search of five different catalogs at least and we have about 22 total across the university campus. So our challenge was if we're really serious about teaching with collections, if you want to take that to the next level, there's a number of things we need to do. We need different kinds of classrooms. We need transportation. We need registrars. We need to do all that, but we need to make our collections discoverable in a single place so that the relationships can be understood and connected or so that the user can find new connections that we didn't even know it's there. And that was the goal of this project. Now this isn't the first time we've tried this. Back in 2009, there was a presentation here at CNI by a former colleague, Meg Bellinger and she was in charge of the Office of Digital Assets and Infrastructure. What was created ended up being a view find based catalog that had about 1.5 million records in it and it sort of languished. And one of the reasons if we look back, why did it languish is because the project and the initiative was technology driven. At no point did we really take a step back and say, well, what's our culture? What's our governance? How are we going to make decisions? Who makes those decisions? There was also a sense that it was neither top down nor bottom up, but it felt sort of imposed from the sides onto the collections. And so there was sort of this bristling of wait a minute, these are my collections. I want to be a part of this project, but instead it felt like those outside of the collections were making the decisions, doing the programming and all of that. And so it was successful in that it created something but that creation was never embedded into our organizations and was never really used. So now we're trying this again. And the question is what's changed and what will be the success this time around we hope. The first is that my position as a deputy and now vice provost for collections was created in 2016. So that means within the provost office, there is someone whose job is to focus on the opportunities for the museums and libraries to work together, to collaborate and to find ways to foster those opportunities. It is also the first time that you have someone in the provost office who knows collections go back to my 20 plus years in libraries. So to have that appreciation and the advocacy in the provost office for someone who really understands what's under the hood as opposed to just a general appreciation for the objects themselves. So that's one thing that happened. Then in 2017, Yale was fortunate to recruit John Barden as our CIO. When John came and he really looked at the landscape at Yale, what he saw was there's a lot of going on within the collections. There's a lot of technology there. We spent an enormous amount of money there. Shouldn't there be something within the IT infrastructure and budget and organization that reflects the importance of it? And he created the cultural heritage pillar in IT. That means we have dedicated colleagues in IT who are focused on collections. It also gives us a pathway to central funding. Every year there's this large project budget that is put forward and we have a pathway to petition and sort of pitch ideas for collaborative projects within cultural heritage. And being able to sort of grab for that money, that was an impetus for us to get together and say what could we do together? Because we'll have a better chance at getting some of that money if we did that. Another thing that happened, which was important, was between 2018 and 2020, three of the four leaders of those four organizations changed over. And that meant when we recruited a new library director or a new gallery or museum director, we could say to them, here at Yale, we have this ambition. This ambition is for collaboration across the collections, not just this project, collaboration at all sorts of levels. And if you come here, we're expecting you to be a leader in that process. So there's an expectation of what their mindset would be coming into the institution. And what that meant is as those leaders came in, they were bought in already, and then they helped to change the culture of their organizations underneath. So as we got together and said, how do we wanna work together? What do we wanna focus on? It was easy for us to jointly commit to IIIF, because if you look at the early founders of IIIF, Yale's one of them. And the other thing that we quickly centered on is we wanted to do this cross-collection discovery project. So we had sort of IIIF built in there, and then let's figure out cross-collection discovery. The commitment is so large and significant that if you look across those different units, as well as University IT, you see money flowing in from their own budgets. You see new FTEs, new positions. Staff positions have been created, and in creating that, it was a trade-off to potential staff positions within your unit. So there's a commitment there. But the largest contribution by far is well over 100 staff have provided in-kind contributions to make this possible. And with 100 people out of 800 total within our cultural heritage units, you have a real commitment overall and an investment in it because so many people have touched this project. So many people have contributed to it and they wanna see its success. It's hard to bring everyone together. It's harder to sort of move fast, but you do it and it's sustainable. Another key part of this is that we were awarded a grant by the Andrew W. Mellon Foundation from 2020 to 2022, extending into 2023. And that grant allowed us particularly to focus on our metadata because if you imagine bringing together those very disparate collections, what does the metadata look like? And it brought us Rob Sanderson, who has just been a fantastic part of this project. Not only does he bring the expertise, but his generosity of teaching others causes people to wanna come and hang out with Rob, right? And so it's like, let's go hang out with Rob. He's working on this great project. And people are just anxious and eager to do so. And we have other stars throughout our project who are just so generous in teaching others. So it's a teaching opportunity for everyone. So what's been built up over time? A couple principles in our culture. One is that all units retain ownership of their data. The day you say, we have to decide how we're gonna have that particular name entry, is it gonna be Trumbull comma John or John Trumbull, that's where it all falls apart because then someone has to trade something up. We're able to retain all of our original systems of record. And even the Lex project, and this is called the Lex project, which is light, does not create a new system of record. So you're not battling who's got it right. We've all agreed that good enough is better than perfect that we will experiment and we'll move forward incrementally. We've recognized that the expertise exists throughout our organization in those 100 people and there's ways for those people and more to contribute to the project. So we're tapping not only IT specialists, metadata specialists, we have curators, we have librarians, we have user interface designers, we have graphic designers, we have university council and copyright lawyers working with us on the project. So you have lots and lots of people and their expertise coming forward. And so essentially what we're finding is that this collaboration is taking us farther, perhaps not as fast because we could have done it if just a small group worked on it but we wouldn't have had the buy-in on the project. And collaboration is rewarded by our university. The dollars are flowing to us because it's a collaborative project and they believe by funding this in the collaborative sense then the individual units will not come asking for a similar type of project that just they would benefit from. Our process is to build consensus through engagement and experimentation as I sort of referenced. We have an exciting vision. We have leadership at all levels. We have really good project management happening throughout. We've established our shared priorities and our values and some of those values include it's an open project. We are sharing absolutely everything we can that is not proprietary. Another value is that as we work with the metadata we're really focused on bias and racism that may be embedded into our metadata from centuries of this kind of work and that we should collectively try to address that. We have an agreed upon governance model and cost sharing. So when a bill comes forward we're not suddenly all arguing who pays it. We've already established how things are going to get paid for and by whom. And if you can imagine this project has many, many teams many groups and committees have formed and each of those starts with the establishment of norms for the collaboration within that committee and it sort of rolls up and through the hierarchy of all the different committees. So all of this to say we pay a lot of attention to how we do things, not just what we're doing. So the results have been really a cultural transformation for us. This collaboration is contagious. We are working on shared collection storage. We'll have the first shared painting storage where the paintings that are in storage from across our museums will come together. We'll have a shared three-dimensional object storage. We have shared positions. First Native American curator for both the art gallery and the Peabody Museum. We have shared conservation labs. We have a shared digital preservation infrastructure. We're sharing provenance expertise across the organizations. We have shared diversity, equity, and inclusion plans. So over and over again, as we sort of tackle something we are defaulting now to why don't we do this together? That has led to lots of peer connections forming across the organization. For reasons that are still baffling me the curators on one side of the street and one museum were not talking to the curators and the other side of the street at the other museum. Literally it was just a street. But organizationally, culturally they were told we are different and we're different enough that we should not be collaborating. We have shown actually there's so much similarity. Why would you not talk to colleagues who could help you think through a particular problem? And over time it's built trust and that trust has led itself to generosity. We now have meetings where amongst the museum directors and library directors and I we've got a problem and someone says well let me put some money down. Let me see if I can kickstart this and the rest of you can come back in. Or let's try to pilot a classroom where we bring collections together. The Peabody Museum said, I'm building five new classrooms. I'll dedicate one of them for that pilot project. Let's experiment and see what happens. So we have this generosity that is coming through that I think is going to lend us to more and more types of projects. So let me turn it over to the more complicated stuff. Thank you, Susan. So, yes, with such a buildup, I hope to at least continue the good work. So, yeah, in terms of some of the technical principles, one of the core ones is usability. Came up in the last session in this section about accessibility and usability. So the data in Lux is linked open data. It's a graph, a knowledge graph, but at the same time it's usable. It's not loud, it's loud. So linked open usable data. And the standards as part of that linked data are critical because with three different domains, with many different units and departments, each with their own cataloging traditions and then changing cataloging traditions over time, we need to ensure that the descriptions are semantically interoperable. We need to know that when we're talking about Trumbull comma John, we're also talking about John Trumbull, we're also talking about John Trumbull dates and so on and so forth. We use an ontology called linked art, which comes from the museum world. It's an international standards and ICOM, the International Council of Museums. We also, of course, use AAAF. But when I say usable data, who do I mean? Who is it usable for? Because usability is determined by use, it's determined by the audience. So instead of being usability of the researcher at the end, instead what we're talking about for the data is usable by the developers, by Jeff's team, Jeff was sitting in the front row. He's the manager for Wave Jeff. The manager for the team in ICS who does all the hard work in setting up the Amazon instances and installing all the software, building the front end, and so on and so forth. So it is Jeff's team that the data needs to be usable for and where are they from? ICS, they're not domain specialists, right? Most of them had never seen before they came to Jeff's team, I imagine, a mark record or any ADSGML file or the headaches that are TMS for the museums. So that model is very simple at its base. It is activity-centric rather than object-centric, that's the principle distinction that we try to draw. So in the middle there is the activity, that could be the production of an object, that could be provenance, it could be an exhibition, it could be the destruction of an object. It's carried out by some actor, it takes place at some location, it involves objects, be they physical objects, paintings on the walls in the museum, or information objects, the works that are carried by the books in the library. It happens in time, in a time span, and it happens in a certain way, a technique, if you will. And then we just take that, that star, and stamp it over and over again across the board with different and subtly different relationships in between the resources to ensure that we've captured the semantic nature of the work. Cliff mentioned fair already, saying, well, hey, it sure would be nice if it was practical. So we have tried to define what we mean by fair in terms of lux, which is findable, accessible, interoperable, and reusable. So in particular, we make every attempt to have our identifiers persistent and de-referenceable. If there is a URI in the system, you can go to it, it will tell you what it identifies. With all of the great cultural changes, a cultural transformation, this leads to also a sustainable system in which those identifiers are more likely to be persistent. As Kuntzi says, persistence is a service. It is not a policy. You can have all the policies in the world, but if you don't have a service, you don't have an identifier. Already mentioned, we use open licenses as much as possible, standards, and keep the data available as possible. I was fortunate enough to give the plenary session, the keynote, for the April 2020 CNI, which was online at the last minute, as you could recall, at which I introduced this equally un-Google-able acronym, Shared. So sustainable, we talked about it, harvestable. So by harvestable, we mean you can go and get all of the changed data since the last time you came to the system. So we want to make sure that it's not just LUX that we're building, not one application. We're enabling other applications to be built at the same time. Available, reconciled. So reconciled ensures that when John Trumbull comes from one system and from a second system, they are merged together. But the more critical point is we also align them, not internally, but also externally, with the library code name authority file, with the Getty Union list of artist names, with VF from OCLC. And from those external data sources, we can enrich our understanding of those entities and make the user experience significantly better. Diverse, we make absolutely every effort at inclusion and diversity. I co-chair along with my colleague, Elizabeth, the bias and racism committee. Governance, Susan is the sponsor. I'm the product owner. Jeff is the person who doesn't actually have a name. I'm afraid. Sit where the strategy thing is. Sarah Prown is our project manager. So we've tried to make every effort to ensure that we have all of the roles in place, to ensure the sustainability and availability of the system. Technical design and process. So it's very similar to our cultural processes. We use incremental prototypes. We make sure that we can always fall back to the previous prototype in case something goes wrong. We try to ensure that our concerns are separated, not just arbitrarily, but through standards to ensure that there is something that both sides of the equation can point to. So linked open data and to play fundamental. We looked for, for a long time, a solution architecture that would help us to build this. And that was also very collaborative. We had about four months of seven or eight people from around the division, both in ITS and from the collections investigating seven or eight different possible platforms. We eventually made a decision that allows for full text search, graph search, and full text plus graph search. This was our core requirement that we had. And I will tempt the demo gods and try, so everyone stop using internet please. We might be interested in radio packet networks. What? Who would be interested in radio packet networks? Celia, who would be interested in radio packet networks? Here's the book in the library. This is our work. Oh, it's LeFiddler. Who? I think everyone knows this guy, right? With his trademark blue shirt. So we went from a work record to a person record. The person records in places have the left-hand side is designed to bring you back into the collection. The right-hand side is designed to tell you about the person. So I've launched a coalition of network information, a member of the coalition of network information. So we do not have the CNI logo in any of our systems. We do not, in fact, have that cliff as a member of CNI. However, this is where the enrichment comes in. So we started from a work of cliffs, we went to cliff, and now we can start to follow our nose and interact with the information that we have built up across the collection of lots of different records. So we can then see places. Here we are in the District of Columbia with a map and so on. And now we can start to see here on the left-hand side more interactions across the different collections. So we have a photograph probably in the art gallery. There are works created at, so from the library, works about Washington, D.C., an image in this case, all works published at D.C. So if we scroll down, we could also explore up and down the hierarchy, see related locations, related agents and so forth. But if we want to go back into the collection, we can then turn that view into a search to then continue to refine and explore. So yes, now we're a long way from packet networks. So we have facets, as you might expect. But remember, all of this is based on linked open data, rather than a traditional solar index. So all of these names are being pulled in from their records, rather than being embedded in this record. Let me point out this one, ecological zones. Because this is an interesting side effect of having multiple domains being reconciled automatically. What it actually is, is landscapes. So in the natural history world, you would say that a landscape is an ecological zone. A landscape, of course, for the painting perspective, is definitely not an ecological zone. But if you go there to the ecological zone record, it will tell you what it is, including landscapes, wilderness area and so forth. And again, it brings you back into the collection to see further records. In fact, we have 1,319 ecological zones, or indeed paintings, sorry, landscapes, from the various different collections. So, let's just look briefly at one more person. So we can also have, related agents, sometimes they're only, let me just go back. Very briefly, search for trumbles, things he was mentioned. So trumble, his objects, works by or about him, the various trumbles, people. Trumble as a place in Connecticut, concepts that involve trumble and exhibitions. You can go through to look at trumble. And for the more extensive artists and so on, we can have timelines of their work. The locations will be plotted on a map, but it's still being worked on. Related agents, trumble is very famous for his paintings of the American Revolution. So here is agents related to trumble and the works of trumble. And we shall leave it there. So, thank you Demo Gods for working. So, some observations about Lux generally. Lux is a knowledge graph, as you've seen. You can very easily go from radio packet networks to John Trumble and paintings and landscapes. Everything gets connected. This is part of the reconciliation and enrichment process. The particular expected search features are available with full tech search, with facets, with connections. But there's also this graph that underlies the data. And we can leverage that graph using the underlying database. So it starts to shift the paradigm from known item search, you go to your catalogue, you search for the thing that you know. And instead it tries to drive the user towards browsing and discovery. So it tries to ensure that the user can leverage their curiosity and explore things that they may not have even known existed. How do you get from the lithographs on one side of the road to the lithographs on the other side? How do you get to the actual specimens in the Peabody? How do you get to the diaries in the library, in the special collections? It's because of curiosity. That's what we really want to drive towards. And through doing that, we think we are bringing it from search into research. Now you are looting things. You are able to investigate the things that you're interested in. For this reconciliation is absolutely critical, and we are so thankful to the Mellon Foundation for their grant. Because if we had five different entities, all of which were John Trumbull, then you wouldn't have those connections. There would be one record from the library, one record from the gallery, one record from the centre, and so on and so forth. Interestingly, we've found that the reconciled records show up more often in relevance-ranked searches because they've been enriched. There's more connections to them, so they rank higher. There's more keywords, there's more descriptions, so they rank higher. So by using the graph, we enable the user to find more graph. It's a virtuous circle. Cross-domain is hard. That's my third observation. I'm sure it is of no surprise to anybody that the paradigms of describing a work versus describing a archival collection versus describing an object on the wall have taken quite a lot of effort to marry together. Also, labels. So in a pre-linked open data world, we use strings to identify things. Whereas in link data, there's URIs that identify this for them. It has meant a mental shift from people just saying, you know, landscapes. Oh, no, what we really mean is this concept. And if the P-Wadi calls it ecological zones and the YCBA and the art gallery call it landscapes. Okay, that was a bad example. It's over-reconciled, but you see the point. Resources are needed. Trailblazing is quite costly, both in terms of technology. Ask Jeff afterwards how much ITS pays for the four duplicate systems in AWS. But also staff time. As Susan said, there's been more than 100 people working on this project over the last three or four years. Some of whom have full-time on it, some of whom only part-time. It is costly in terms of time. We have chosen to do this. We could have chosen to do something else. There is a finite number of people that work at Yale and I'm very grateful that Yale has decided to put its money where my mouth is and it comes to linked open data and this sort of thing. Hopefully, you've seen how it's going. So, that said, there are a lot of flowers along the way as per the photograph. There have been many joyous moments as well as many head-scratching moments as to why the technology is not working as we anticipated. So, to start to wrap up, we have some known unknowns in particular the extent of adoption. So, we believe that this is going to be well-used and well-loved but it's all in development as you saw. There are some rough edges. There are some visualizations to be finished. I did not demonstrate advanced search because it is not quite ready yet but we will be launching in spring. So, one of the questions that we've tried to keep in mind is what is something that is intuitive and innovative because it's very hard to find that intersection but that's where we're aiming. And then sustainability. By trying to ground this in teaching with collections, by grounding it in the cultural heritage IT pillar, by ensuring that all of the units are participating, we believe that we can make the sort of work mission critical but we don't know. We've seen the ODI if at Flounder, we do not want to see this happen again. So, one of our next steps is instead of using the data set and the platform just to surface this particular application, LOX, instead we want to have it as a service whereby the units or anyone can build their own applications on top. We've already seen some of the units say hey, the data in LOX is better than the data in our collection management system. Can we come and use the LOX data as a new collection website just for us? Because of this enrichment process, because of the reconciliation, because of the connections to all of the other resources. So, we are hopeful that by embedding the work within the cultural context and the technical context of the university, that it will be sustainable. So, Susan mentioned that we will be sharing as much as possible, that's absolutely true. We do use a proprietary database engine called MarkLogic, which is here in DC, which gives us that graph plus text functionality. That said, while we can't share that code, we can share everything else, including all of our data and image content, the processing tools for the data, the configurations and applications that were built on top of MarkLogic, the technical documentation and the documentation. If there's something else that you would like us to share, please do ask. It's only because of a lack of knowledge on our part, that we don't know that this would be useful and that we're not sharing it. With that, I will hand back to Susan. Just to summarize, this digital project is helping to shift our culture, is helping to change our culture. As Rob said, we've moved from collection discovery to really thinking about how we're going to change our culture, how we're going to change our image platform. We're improving our collection data and hopefully other people's data as well. We're bringing others along. It's improved our understanding of our internal culture, but more importantly, it's causing us to ask hard questions of the originating cultures of the objects that brings together these diverse opinions across the room to say, should we be doing this or should we be doing it differently? It has caused a more informed understanding of our cataloging practices from across the collections. You have those who catalog natural history objects talking to those who catalog books and there's knowledge going back and forth. There's a growing understanding of the different domains that are there. That people are really coming to want to work with one another. Early in this project, when we said we wanted to try to build this on the idea of linked open data, we would say, well, what's linked data? I don't understand it. It seems too hard. It seems really complicated. It's adding time. Why are we bothering? A colleague who just recently said, the entry for N. Houdian does not contain her name in cuneiform. Can't we link to that? Suddenly, we have curators and others who are just thinking about linking is the solution to lots and lots of problems. Why just be left with what's on the page here? Let's keep going. That's indicative of the kinds of new conversations that were happening. With that, I think Rob and I would be very glad to answer any questions or have a discussion. If you do have a question, please do use the mic so that it can be recorded. You said you're using a proprietary database behind it for open-link data. Why did you use it for proprietary? Instead of some of the other technologies? As part of the research process while selecting the backend platform, we had two teams of four people each each trying to build an architecture that pulled together open-source products or AWS products to do the same thing. They were Neptune for the Amazon side there was Neptune, Elastic Search and DynamoDB and in the open-source space there was Neo4j, Solar and MongoDB. We concluded by the end of probably the first half of a four or five month research exercise that the cost to maintain those connections, maintain the integration of the individual platforms was going to be much higher than the not cheap cost of getting a proprietary system. We're doing this because it gives us great flexibility in terms of how we can construct the queries. It has several internal APIs for building hybrid graph and text queries which has been absolutely invaluable in terms of advanced search and building all of those connections you saw in the demo from the individual two pages. It was absolutely not my choice. I'm an open-source guy by heart. I think a lot of us are but when we saw the performance and there's 50 million records 861 gigabytes of JSON-LD in there. When we saw the performance and the flexibility we were sold. I should say we are a not-for-profit. We are a not-for-profit by heart. I'm Sebastian Hammer with index data. Thanks for your presentation. I was curious, you mentioned early on that one of the criteria for this project has been to avoid creating a new system of record. It suggests an ongoing ETL type process out of the existing systems. starting to see the value in creating links and embracing this. So, do you see a path, a roadmap for moving or pushing this link data approach out into some of these legacy systems? How do you see those playing in the future? And I suppose specifically, I'm curious about whether some of the work that's already going on with link data and the library community, does that help this? Do you see it as kind of being... I know that you've sat in and been part of some of the link data for libraries, link data for production types of projects. Do you see that path and that work complementing this, or do you see this being a different path, a different perspective on this? No, I think they're extremely complementary. So, the way... So, yes, you're absolutely correct. There's an ETL that harvests the records nightly from the units and puts them through a pipeline that does the reconciliation and enrichment and transforms the records from Mark or from TMS data into link open data. So, the first step in that pipeline is owned by the units. I should be very clear. So, we, ITS and the Cultural Heritage Division, generally, do not own it centrally. Each individual unit has their own ITSARF, and they are responsible for maintaining their connection into the system. That pipeline then uses the $0s in $1s from Mark Records. It uses the equivalent ALSAIMAS links or the close matches to start to bootstrap the connections. Then, once it's got a critical mass of connections, it can then also use traditional label-based reconciliation. But it relies very heavily on the outcome of the Mellon Grant, whereby we had students and staff working together to find that critical mass of identifiers and import them into the systems of record in the units. We're short on time, but there's one more question. Yeah, Annie Murray from Calgary. Really hoping you'll come back and tell us how this impacted your teaching culture. So, if the people didn't cross the street to talk to each other then, but now the system is crossing the street, how do you think this is going to change the teaching practices of your various specialists? And will you make any physical accommodations or movement of objects around campus to enable that for the, so the students see it all? Thanks. Yes, to both. So, for example, one of the many committees that have sort of spun out of this is a Teaching with Collections Committee, and they are educators from across all the different units coming together. And we also have a Teaching and Learning Center, and we're working right now to embed a collection specialist into the Teaching and Learning Center so that when faculty come in and say, I want to consider different pedagogies for my classroom, there might actually be someone there who can then sort of refer back to us. So, it is, has sort of spread to that point. We have shared collection storage spaces that we are building so that you can sort of start to come together. But to your point, in order to teach across the collections, we're going to have to agree on security levels, transportation, art handling, temperature and humidity in the room and go on and on and on, but we're committed to doing it. So, I briefly mentioned that when the Peabody Museum, which is going through a renovation right now, reopens, next fall we're going to dedicate one classroom, which is in the secured area, to start exploring, what does it take to get all those pieces together? We do imagine a future where rather than having art handlers embedded in each unit, we start to create different positions that cut across so that it's one delivery truck and not four delivery trucks moving things all around. But there's a lot to go on that process. So, thank you very much. I know we're at time, but we'd be glad to answer any questions because we do end at five, right? Yes. Glad to answer any questions that you might have and we'll see you also at the reception. You'll be there? Yeah, we'll be there as well. Thank you. Thank you.