 So this is the mapping of the CRS terms. Now the terms themselves are publicly available and David from Archives showed an expression of them online. And behind the scenes the terms are listed in a relation database. And you don't just have the terms, you've got the individual terms and also terms that those terms are related to and the type of relationships. So these are semantic relationships right here. And so, Nagan, this is an expression of exactly what you were asking about, not in RDF or anything like that. On the right column where you see NT, that's a narrower than match. Where you see EQ, that's an equals match. Now these are local terms being used within the CRS system, but we can map them out to RDF and that's exactly what we've done. So the process goes something like this. We just do a few transformations in Excel to get unique IDs and neatly printable labels for these terms. So that's straight just column transformations. Then we take the actual semantic relations in their system and there's a bit of discussion between David and myself as to exactly what it was meant. Now NT clearly means narrower than or narrower in SCOS, but a few of the other matches require a bit of understanding to work out what the SCOS equivalent terms and mappings were. And I was very unkeen to invent new semantic relations if possible, I would reuse existing ones. So here you see a SCOS relation and in the particular term IDs. Again, these are just column transformations straight out of the Excel. And on the right column there, the full relation, that's just concatenating bits and pieces from various columns, but it's approaching an RDF format right there. And then finally, I extract that column and a few other bits and pieces and actually create an RDF file. And I take that RDF file and I put it through a validator just to check that I've formatted it correctly. And then that validator also reformats it kind of neatly. One thing that's important to know is that because of the way RDF and graph systems work, if you've got all of your information in Excel or some other tabular system, you can assert each property of terms and each term relation just in series, one after the other after the other. Put that all into your final Excel column, put into text, throw that into an RDF graph system, and it will normalize and join everything up. So you don't have to spend forever. You've got to get the identity of all the elements and joins in Excel, but you don't have to yourself join it up if the word matches the graph database systems will do all the joining. So this is an easy one because the terms in the CRS system already had all the semantic information. In other systems, that information isn't there and other work has to be done. But this is one of the easy ones. Oh, so this is a different piece of information. It's actually core to the project, but I didn't have time, I thought, to present it in the main presentation. It's really just handling the changes that we experienced in the project. So we can have changing government entities. So here the diagram is just showing a theoretical split. So agency X gets split into agencies Y and Z. And the way we deal with this kind of change is just to have these change events able to be recorded. Now the National Archives database already does this with change events and proceeding and succeeding agencies. And we're not sure if the eight-course system at the Department of Finance tracks all this information. If it doesn't, we're going to have a hell of a job back working out all those change events. But nevertheless, we can think of all the ways in which agencies change. And we can describe that change through a series of change events. So that handles changing organizations or entities. When concepts change, it's a bit different. Within the vocabaries, we don't have too many examples of this. Within a vocabary, but certainly across vocabaries, we've got departments that are associated with functions. So here you've got department X associated with function Y. But if function Y gets superseded by function Z, then of course we can infer that now department X is associated with function Z. And we have to know whether Y and Z are equivalent so that back in time, we can say the department was associated with functions Z. Or is it going forwards only? So those are kind of the nuances there. But in general, we can handle changing concepts within vocabaries themselves. And as David mentioned, having a split between the organizations and the functions and the organizations in his case and the record series allows you to sort of handle changing all those environments. But now there's the really tricky ones. The last slide I've got, it's managing change between entity function mappings. So we've got an entity that itself may or may not have changed. Let's say it hasn't changed, so department X. And the functions themselves haven't changed, but the mapping has. So department X has acquired a new function. And so the way we deal with that is we say, in a simple way, we would say department X is associated with function Y. That's not enough information to handle entity function mappings. So then we say, look, let's take that relation out and talk about information about that relation. And so in the next, the more information section I've got there, the subject of this relation is department X. The particular kind of relation is the associated function one. So that's the predicate. The object here is function Y. So the first three lines of that more information section are exactly the same as the top assertion, department X associated with function Y. However, we can say anything else we want about it. Now here, I've just got created at time. So when was this assertion created at time A? And then perhaps this assertion was invalidated at time B. So maybe it was, maybe department X was associated with function Y in 2014. And then due to machinery of government changes or something, it was unassociated with that function in 2018. So you'd have created time at 16 and invalidated time at 18. So now we've got a time bound association and we could put any other information in there. Now of course the departments themselves can be changed and the functions can be changed. So you can see with these three techniques we can handle an awful lot of change. And it's no different to what databases with different elements like the CRS already doing. But here we have a very powerful, extensible method. We can map across vocabaries. We can change within the vocabaries. We can map across datasets. I haven't actually talked too much about how you would map a department in one to an agency in the other, but you can imagine it's similar. And then here's this general technique of talking about assertion relations. And link sets do this. Link sets say, look, here's an association I'm making. And if we want to, we can record any amount of information about that assertion like times as you see here, but also things like what logic or what method was used to actually create this assertion. So if it's a human asserted thing between two vocab return, the method might be David Herder's expertise. But if it's a statistical mapping, we might say the method here is emergent statistics from some kind of analysis. So that's really the end of my presentation there.