 I would like to thank Nathan for inviting me to speak and just to explain we're going to split this talk into two parts. I'm going to say a little bit about specifically what I've been doing with regard to data on the Beyond Boundaries project, but the last time that I wrote a paper on data management and databases in numismatics was five years ago, so all of my knowledge is now obsolete as dinosaurs. So my colleague Amelia Dowler, who currently deals with a whole bunch of databases, SNG and the various ANS Oxford databases, she will take over halfway through the talk and give a bit more detail about how numismatic databases operate more widely than simply the one that I'm working with. So I'll first of all begin by introducing the British Museum's database. Now this is the one that I've primarily worked with during the project. I've worked with other numismatic databases in the past, but the BM1 is the one that I'm currently spending most of my time working with. And somebody in our AS department will get irritable, I'm sure, at the simplifications in the description I'm about to give you, but it consists of four different databases. A database for managing the metadata associated with the objects, which was once called MAGUS, is now, when I arrived at the museum, was called Merlin and is now called MI+. A database currently called Odin, which manages all of the images associated with objects, these are the two databases that we put data into. So upload images into Odin, type things into Merlin, then connect those in some way. And then two shadow databases that exist within the museum that disseminate this. So a copy is made of MI+, and a copy is made of Odin, and collectively these two other databases are called Collections Online. And if you access data from outside, so here's our input, here is you accessing data from outside, what you see are these two shadow databases, which in theory contain the same information as the two places where we put data in, accepting, of course, things like personal addresses, precisely where the object is located in the museum, the kind of thing that we don't want you to know, but we want to remember ourselves. In principle, in practice, the link between those has been broken for how long? That's an unfair question since July. Since July, okay. So what you see when you come to the museum is not the data that we currently see when we're looking at this. The system is excellent at preserving data. It has mixed success with disseminating data. So what it effectively acts like is a repository. Now, it's worth saying that alongside these two databases online, one of the things that happens is to transform the data into a form that other computers can get at. And Amelia will talk a little bit about portal databases in a little while. Every department in the museum has a different workflow, a different way of arranging what it does. Coins and medals, getting data onto the databases consists of three steps. You have a pile of things and you have to put them in order. And you must do this before you can do anything else because the data that we put on is not inherent in the objects. It's created by our taxonomic processes. It exists. A really good example of this is diameter. Ancient coins do not have a diameter. Circles have diameters and ancient coins are not circles. But every coin that goes into the database gets a diameter, which means you have to decide which measurement you are going to take and how you are going to take that measurement, rulers, calipers, before you can actually put data in. And this is not consistent. I measure the diameter of a coin from the top to the bottom of the design. It turns out that the colleague that I worked with five years ago entering the same set of data, alternating coins, measured whichever distance passed through the centre that was longest. And my Roman colleagues downstairs all measured from left to right. Meaning three different measurements were going into the data. Once you've got everything in the department ordered, you can create metadata, the diameter and description of the coin, a type. A hugely important typology that you've created academically can go into the database that allows the coins to be grouped and recovered later. And you can take images of the coins. Ideally images of coins consist of two images welded together. The front and the back of the coin placed into a single file so that they're not easily separated. One of the issues that arises with this workflow is time. It's very, very hard to automate data entry for coins because it requires you to know something about coins. It turns out it's incredibly easy to automate photography. When I arrived at the museum, the practice was to scan coins on a flatbed scanner. Then take the image you had scanned and use the white paint brush tool. Everybody is familiar with the paint brush tool in Photoshop. And colour everything that wasn't the coin white by hand. Very carefully so you didn't overrun the edge of the coin. Now we put coins underneath a fixed mount camera. We photographed them under a vertical fixed lighting that doesn't cast a shadow. And we have the computer run over lunch, remove all of the background information and combine the obverse and reverse image together so that it's all done when we get back. It takes about 30 seconds whereas entering the same, even basic data for the same coin takes about 5 minutes, significantly longer for certain coins than that. Exactly how long it takes to order the coins of course depends on how much people already know about them and how much of a mess they're in when you arrive. So there are differences in the amount of time. And this creates an issue with our activity. So for example, we held an event on the Huns recently. As part of that I spent 5 hours and I photographed every relevant Hunnit coin in the department. Which meant that I could pull them out when I needed to for the event, I could carry them with me to the event. To have created data records for them and remember we need both to get them online would have taken about 4 weeks. So that hasn't been done. And it did take me actually until relatively late in the project to realise that A, that was really bad and B, there was a way available in the project to solve this. The issue is that the database itself is very robust. It's backed up extensively and off-site short of some major natural disaster striking the whole of South England. The British Museum's data is going to survive. However, the images of coins are stored on memory sticks, CDs and portable drives. They wouldn't take anything more than somebody accidentally leaving a magnet in the wrong place to get rid of everything you've produced. But we do have Xenodo. We can upload all of the images while they're waiting for their metadata to Xenodo and store them safely there. There are some limitations. Xenodo is even worse at disseminating data in many ways than the British Museum's database. Once they're up there, you don't seem to be able to pull all of the images back out. Unless you also upload a second copy. You can't change them if one of them turns out to be the wrong image. It's nowhere near as good as the British Museum's database. But what it does is temporarily store, or in fact permanently store, this part of the workflow. Here are all the images now. At some point in the future we might get round to being able to put them on the database. One of the problems that databases have and they're much more apparent at the British Museum than anywhere else because the British Museum is very old. So it got into databases in the 1980s for recording its collection. The ancestors of MI Plus are nearly 30 years old now. In fact probably slightly more than 30 years old now. So we actually experience conceptual and linguistic drift in our database. When we first introduced the field that records an inscription, somebody decided that the appropriate way to do this was to have a field within it for the language and a field for the script. This is a drop-down from the recent language field. You will see that there are seven, eight different entries for Cypriot Greeks of one type or another. There's some interesting ones on there that look... Dunish. I'm not sure if Dunish is a typo for Danish or if we genuinely have some Tolkien-related artefact in the museum. D and E, I've no idea what somebody was thinking when they put that in, and Devan Agari as language. But my favourite of all the things that got into this field over the years was, does not seem to be Latin, right? Which some classicist clearly believed was all the information you needed if your inscription was not actually Latin. So what we need are guard dogs. This is something that we have at the British Museum. We have two features to the database that make the database particularly ideally suited for numismatic data. Other than its robustness, one of them is multi-user curation. I am curating data that other people in the past have created. Other people will curate it after me. Other people are checking me. And more importantly we have guard dogs. And the guard dogs check that we're all using as far as is possible roughly the same concepts when we put our data in. And one of them at the moment is running through that language list item by item making sure that everything in it actually makes some coherent sense. Amelia is going to talk a little bit more about that in a second. So the longevity of the database, and we now have practical experience in that because the database is 30 years old. But we want it to be 200 years old. It's going to need to be there in two centuries that data is the robustness of the storage, the multi-user curation and the guardianship custodianship that's exercised over it. What the database is still not good at is version control. Though it's getting better, right? And dissemination. But in a sense the dissemination is secondary to having the data stored in a usable format. And Amelia will now explain some of the ways in which other numismatic entities have begun to engage with the problem of dissemination and some of the practical issues that that has brought on. Thank you for joining us. Yes, I'll reach the computer. Otherwise I'll have to get you to change slides for me. Oh, yes, that's all. It seems a bit odd. Hello, everyone, and apologies. You may have noticed me scooting in rather late. Please blame southeastern trains and not me. As Robert said earlier, I'm going to talk quite broadly about numismatic databases. I don't know if there are any numismatists in the audience. I'm assuming not. But what I say is applicable to any kind of specialised object database. The same sorts of conceptual problems occur whenever you're dealing with a specialist object. Thank you very much, Robert, for inviting me. And Nathan as well. I haven't met you. You are Nathan. Thanks, Nathan. Again, sorry for being late. So what I'm doing here is a very brief survey of how institutions deal with their numismatic data and also a little on how projects deal with numismatic data. These are all object-based databases. I have a limited time. This is necessarily quite brief. And this is a developing field. So for a lot of what I'm going to say, it's developing. People are still working on this. So to start off with, I've brought up two examples here. On the left you can see the Bibliotech National in Paris. They're Gallica websites. And over here, this is information about the Ashmolean's collections online. So this first part of the talk will be about institutional databases, how institutions deal with their data and what you may not appreciate about what the institutions are doing behind the scenes and the effect that that has on the presentation of their data through their databases. So on the left is the Gallica database from Paris. As you might guess from the name Bibliotech National, this is the National Library of France. Coins, as objects, have had a peculiar history in many ways and were quite often classified as manuscripts and thought of as metal manuscripts because of course many have inscriptions. And that is the history of coins in the British Museum as well. They used to be part of the manuscripts departments. And in a number of major institutions around the world, coin collections are still held in national libraries. The coin departments in Bibliotech National came across a problem a few years ago. Their building had to be redeveloped. They were going to close their collection to all researchers and people would not be able to have access to their material for a number of years. The solution to this of course was to digitise as much as possible so that during the renovations people would still be able to access the collections. However, the existing institutional database is a library database and so the coin room was faced with shillhorning their information into that database. You may not be able to see it at the back, but the mint I've given you, an example of the mint of Athens at the bottom here, the mint of Athens is under author in the library database. This is what they had to do to get the information out there. This isn't a permanent solution. They are working very hard behind the scenes to improve this. But this was the quick fix that they came up with to sort out this problem. It does mean that searching can be quite difficult. There isn't very much data on there at the moment. Again, having specialist subfields from the misbatic data, these don't already exist in the library database. So there's a lot of work going on behind the scenes to try to fix that. But for the time being, there is at least a way at getting at the material. So if you didn't know that history, this looks like a very odd choice of how to present your data, but there are reasons for it. The Ashmolean, when they started putting their collection online, followed quite a typical path for institutions, which is that departments within the institution might have had someone with an interest in databases or funding, and they developed individual databases. So here's a list. Actually, that hasn't come out terribly well at all. There are multiple Ashmolean object databases, which were not connected with each other, which had their own fields, their own terminology. So you couldn't search the entire Ashmolean collection. You had to have specialist knowledge and go into, for example, Roam Provincial Coinage, and search for data there, and then come out of that and search the Beasley archive, for example, for further information and try to connect it together. This is, as I said, this is a very typical way that institutions in the past have approached the management of their data. They're publicly facing data, I should say. And so what the Ashmolean is currently doing is working on having a cross-institution database so that you can access all of this material through one search. This is particularly important if you're not so interested in coins. I can't think why. It might not be. But if, for example, you're searching for images of, I don't know, the goddess Artemis, you'd be interested in coins, yes, but you'd be interested in vases, you'd be interested in mosaics. You'd be interested in a whole load of different kinds of objects, and having to search individually is, say, the least annoying, and it runs the risk of you missing certain types of objects. So here are the British Museum and the Ashmolean. This is how the Ashmolean is now presenting the object searches. You can search across the whole institution, but you can still have access to specialised databases. You can restrict what you're searching for. And I think this is actually quite a neat solution. It's a work in progress. They're still adding a lot of data to this, so not all of their institutional data is out there yet. One of the criticisms of the British Museum website, and this is the search page, is that you do search the entire institution, and quite often people come to our website after very specific classes of object or information. And they're faced with this. People find it quite difficult to use, and they may pull back things that they're just not interested in. Now, given that we have several million objects in that database, this can be problematic if you've used quite broad search terms. Even when you're searching more specifically, you can often pull back thousands and thousands of records as well. So this is something that institutions really need to think about how to present this to the public. I, as a member of staff at the British Museum, I could use this, I know how it works. It's great for me because I can do all sorts of detailed searching. A member of the public will find this quite difficult. As Robert highlighted, the collections online of the British Museum has not been updated since July. This is because these conversations are happening behind the scenes, and there is going to be a new way of searching the database soon. We don't quite know what it's going to look like yet, so I can't tell you. But who knows? I'm hoping it'll be much more user-friendly. So, as Robert said, our database has been around in various forms for about 30 years. When you're approaching, as an institution, making a database, your institution has probably been around for a few hundred years. You want your database to be future-proofed. You want people to still be able to use it. What we don't want is every couple of decades someone having to rewrite the whole thing. This is a waste of time, it's a waste of money and effort. I've spent an awful lot of time putting data into the database. I do not want to think that whoever succeeds me in my job will then have to redo absolutely everything. It's wasteful. So, I'll put the top right. This is one of our old access orders. This is actually brilliant because it's got pictures of the coins. Someone's lovingly drawn them in. You can see down there on the right. That's very unusual for coins. Usually it's just a line or so of description. This is not always enough to identify an individual coin. Coins, of course, are mass-produced. So, you might have ten examples of a particular type in your collection and you don't necessarily know which one it is. We don't put object numbers on the coins. They are only identified by the information that is with them on small tickets. So, if you were to drop an entire tray of these things and have to put it as I have had to do in the past, someone else dropped the tray, if I would like to point out. I wasn't that angry. You do have to put the data back together again and it can take quite a long time. This is an example of collections online. This is one of my coins from Carthage. You probably can't read any of it. Don't worry about that because there's not very much data there for you to read. There's a nice big blank square for where the image should be. What we don't want to do on databases is replicate the errors of the past and not to have very much useful data there. However, the decision was made when we released our data online was to release absolutely everything, apart from anything that was security-sensitive. So, you do get records like this. Other institutions take a different approach. The Moons cabinets at the Bodum Museum in Berlin, for example, they have beautiful records. I highly recommend going and having a look at them. You can zoom in on the images of the coins, lovely descriptions, absolutely beautiful, but it is limited to what they have had time to do so far. They have not got their entire collection out there. At least you know how many coins we've got, but there is always this conversation between an institution like mine and my colleagues in Berlin who have this fantastic, useful, wonderful, beautiful data and ours, which looks like we're not trying. I have 100,000 coins to look after. I haven't got to this one yet. So, I'll do my best. The Ashmolean, the Hebron coin room there, are taking a middle ground. So, you have an option of seeing their approved final records, or you can click on a button and see absolutely everything, no matter what state the data is in. But at least you're forewarned that it may not be perfect. Although that does raise the question of, is data ever perfect? Will there never be further changes? For example, if something goes on exhibition somewhere, should there be a public history of that? If an attribution to a mint or a ruler changes, how final is the final data? With everything that you've seen so far, you'll hopefully have noticed that there are quite controlled fields for the data. There are still a number of databases out there. At the top left, this is Coin Archives, which is a archive of auctions, coin auctions. This employs a lot of free text. It can be quite difficult to search and to reduce down to the kind of data you're interested in. But again, at least all the data is out there if you have the time to plod through all of it. It's a very blunt tool, and you can't use it in a very detailed way. So, the trend really is towards these more controlled fields. But, as any institution will say, it all depends on the resource that you have to be able to input good data into those controlled fields. This requires considerable staff time and expense. One of the problems, of course, is what happens when a particular project finishes or a curator leaves or whoever is responsible for that data leaves if there is no replacement. What then happens to the data? The traditional way of publishing coins is over here on the left. This is the earliest coin catalogue from the British Museum, and that was published in 1826. So, there has been a very long history of publication of the collection at the British Museum. It's the same of most major institutions which hold coins or indeed any object. This on the right is the Syllagene-Nymorum brycorum website. The Syllagene-Nymorum, or SNG, project is the oldest British Academy project. It publishes collections of Greek coins in the UK in hard copy catalogue form. The database on the website replicates that hard copy. It never changes. It is a digital version of that hard copy. This can cause some institutional problems because there are a couple of British Museum SNGs on that website. Because it is a much more specific set of search fields for coins, researchers will quite often use that, and then email me asking me questions. But the data may not be, there is a drift between the publication of that data and what we now have in our own database. This can cause problems, and it will only continue as the years go on. This also raises the question, do we need to track the changes? Do we need to track when we change attributions or when people's opinions change of which mint is which, which ruler produced which coin? At the moment on the British Museum database, this is something you can add in a free text field. I haven't seen on any other databases controlled fields where you can put a history of an object. I think increasingly this is something we're going to have to start tackling. The future of databases is probably portal databases using the institutional data that places like the British Museum or the Ashmolean or the Bibliotech National put through. These are two examples from the American Lismatic Society. This is coins of the Roman Empire based on the Roman imperial coinage catalogs and Pella, which is based firstly on the Alexander coinage published by Price in 1991. They are dependent on institutional databases and people there putting in the right data because then that data is extracted and presented in databases like this. These databases have a big advantage for other project databases in that the American Lismatic Society has given institutional backing to these projects. Quite often the problem is when a project ends, the database is orphaned and it becomes obsolete, links break. This can be difficult and there are huge implications for firstly getting the data in lines funding implications for that but very rarely do people work funding for maintenance into their projects or consider what happens to their data afterwards. To finish up, some specific object problems which as researchers you may not think about but this is how institutions approach things. In the world of Greek coins and Lismatic more broadly there is now a project called Lymysma which is an international group trying to come together and explain what terminology we use and so that can be applied broadly across databases. Hopefully this will have been given the stop sign. This will mean that in the future conversations can continue and institutions can provide data for researchers to use which can then be replicated and managed in a more appropriate fashion. Thank you very much.