 Dr. Adam Medusin comes to us from Harvard University, where he presented a dissertation called entitled The Old Syrian Social Network, an analysis based on the text from Kanesh, 1950 to 1750 BCE. He is currently a Mellon postdoctoral fellow in the digital humanities here at UC Berkeley. And he's been working on a number of projects related to social networks. And so if there's a little time after his talk, then you can do a live demo of a related project that he's also working on inside. Thank you, Adam. Thank you. Thank you, Niko. Exciting to be here. I'm so glad to be able to talk with fellow archaeologists. I think that most of you will probably be familiar already with a lot of what I'll talk about. Kultepet Kanesh, it's a fairly well-known site nowadays. But I hope to show kind of a process of how we go from some of the most challenging aspects of archaeology that being unprovenanced artifacts and unknown locations to more statistical and close controlled recontextualizations of these archaeological data sets. So the primary sources I work with come from one peripheral Bronze Age site known as Kultepet in Turkey, which initially yielded more than 5,000 unprovenanced Cappadocian tablets. The majority of these were sold in bazaars and collected by Western museums before World War II. And after decades of textual scholarship by a small group of international scholars, which my work has benefited from, I focused on situating these texts in a reconstructed social and geographical setting, which can be analyzed and described using a network analysis and other computational methods. Archaeology in the Middle East or in the Near East, as it's known for, has long and illustrious history with more than 150 years of scholarship. From the 1840s onward, Western archaeologists like Sir Austin Henry Layard made early discoveries of textual and material artifacts in the heart from the heart of Mesopotamia, and awoke deep curiosity and deciphering the beginnings of human history. Unfortunately, these discoveries inadvertently incentivized a significant amount of looting in the area, resulted in thousands of unprovenance artifacts. When we walk through the museums, the British Museum and the Louvre and the Met, we notice it, right? I mean, it's something that we notice while at the same time not noticing. And so the same goes for this site. And the first horde of texts that were found came from locals. So beginning in the 19th century, a common era, a number of cuneiform tablets emerged in these antiquities markets, first appearing in the bazaars and eventually reaching the major museums of Europe and America. Professor Archibald Henry Seis, the curator of the British Museum, reported one such discovery. In 1881, Dr. Pinches drew attention to two cuneiform tablets, one in the British Museum and the other in the Louvre, as they were said to come from Kayseri. He proposed for them the name Cappadocian. Just before the war, 1,200 tablets, mostly in a perfect condition, were discovered by the peasants, 800 of which were seized by the Turkish government, but the rest found their way into the hands of dealers. Some of them came to Paris and were bought by the Baudelaire Library and Ashmolean Museum at Oxford, as well as by myself. What has become of the others I have failed to learn. So we see that, of course, a lot of these collections ended up in private collections as well as museums. Since the initial discovery, more than 5,000 unprovidence Cappadocian tablets were collected by these museums. After deciphering the script, they were able to determine that these are the remains of an intricate Assyrian trade network, which dated back to 1950 BCE. After the war, Turkish archaeologists were able to resume excavations at the site from 1948 onward. However, due to the dislocated process involved in the preservation and curation of the tablets, it has been extremely difficult to re-establish the location of these initial 5,000 at the site, in particular the rooms in which these archives were kept within the private houses. So for example, one such account from the director of the excavation during, from 1948 until quite recently in 2000, Taksin Uzguc wrote, in his field reports, an archive of 947 tablets and unopened envelopes and pottery was found in two small rooms. They were evidently kept on wooden shelves against the walls. The tablets were found on the floor along the walls. Were those that fell off the shelves in the fire. The tablets had been packed in bags and straw wrappings. And sacks were discovered in piles in the middle of the rooms. A group of the tablets, as usual, were kept in pots. Pottery was set along the base of the walls. Next to the hearth were found unbaked tablets evidently left dry or prepared to be fired. So that's all we have in terms of the real documentation, a number, 947 tablets from a general area where we know they're excavating from the other artifacts that came at the time. Unfortunately, there are no further details as to the tablet numbers, museum numbers that were given to these tablets once they were pulled from the ground, moved away from the site and then located in the museum or anything that would help us link the individual tablets to the rooms which they were found. For example, a house plan number or a grid coordinate or a locus or some kind of stratigraphy. They'll distinguish general stratigraphy from different periods and time in the site but nothing very close, unfortunately. So what their excavation reports do tell us are the few published details about the clues of the Assyrian filing system. So in particular, these tablets were located on the floor above what appeared to be with something above appearing to have looked like broken shelving. And we know from other ancient Mesopotamian sites like Sipar that we find shelving like this and there could be some similarity drawn. This, in fact, is a door from the Ankara Museum and they still seal the tablet room. They literally seal it. They'll break it open, cut it open each day and then at the end of the day, they'll bring it together and press together some sealing over these wires exactly as the Assyrians described their archive rooms as well sealed in the same way, if funny enough. So, and then where they also kept these tablets in pots and sometimes baked in the hearth. Naturally what little we're told only raises more questions. So how did each merchant file their tablets? Because this isn't an administrative center. These are private merchants so focused on long distance trade. Each of them had their own kind of writing style to a certain degree and we can see that in the text. Their hand is evident rather than a scribal tradition that we find in Southern Mesopotamia and Sumer. And do we see any consistent practices within the colony either from the Assyrian or Anatolian households? And that's another thing that this group of texts it helps us to understand is that it's clear from 1948 onward that the excavations have and that the archeologists have kind of pointed to different types of households that have yielded different types of artifacts and these types happen to be either Assyrian which makes up the majority of the site. Sorry this is blurry but this is maybe a little clear. Assyrian with like scattered in between here Anatolian houses as well or so-called Anatolian houses. Now of course the cultural ramifications of all this is lost very often because we're working so late in antiquity but from the texts we've been able to note that at least going off of naming practices which isn't a sure fire way to distinguish cultural or ethnic background but at least using naming practices we can identify four distinct Anatolian naming types. So we find Hurrian, Hittite, Hattie and Louvian names and these names come in somewhat of like clusters in the site as well with the large majority being Assyrian names. So with what little I could use from the archeological side of things the plans and the field reports I then focused on the relationships between the families as they were recorded in the texts. And that in and of itself was what kind of took up the bulk of my research when I was working on this for two years ago. On the basis of close readings it was clear that the initial 5,000 tablets discovered before Turkish excavations came from one general region this region at the site of Kultepe and after making a composite map of the published texts and comparing the houses with the descriptions and what little was known from the excavation reports we could begin to reconstruct the extensive lists of persons and family trees which were conveyed through the texts by aggregating mentions of the same individuals across these 6,000 tablets basically. And what initially became quite clear looking focusing on this site at least was the real prominence of the merchants that were living here and I'll explain kind of how I could make a claim like that in a minute. But at least from the texts themselves I could revolve and able collectively through for the last like 50 years to start to establish family trees that extend across six generations. And by linking the certain archives back to the room numbers where they were presumably excavated we can then hypothesize a certain number of generations that may have lived in the house. So Pushuken is unambiguously mentioned more than any other merchant about 450 times across these texts. And from the networks that I've reconstructed it's clear that his ties run deep in many directions and extending into the Assyrian royal families as well. So indicated by letters to the king and the sons of the king, the prince, the rubatum and branching into the far reaches of the Bronze Age market in the distant hubs of Anishholia where he lived at least part of the time in his life. He died in or at least he may have died in Durhumit which is an unknown location still being trying to be determined by a combination of texts and archeological practices. So no doubt because of this merchant's high profile he was privy to a broader perspective and his advantage as a leader in the trade also extended to his children. From his oldest son Buzazu to his youngest daughter Ahaha who was a gubabtum, a type of priestess in Ashur. She was afforded a house of her own that she was maintaining for the family there and they frequently write letters back and forth. So she's a thousand kilometers distant from this site and yet here in the archive very present and involved especially at these critical moments of death and lawsuits, the lawsuits that ensue when a wealthy merchant, an established merchant dies because all the creditors come in all of a sudden and want to make sure they get theirs. So it's very clear that these family trees are evident from the texts and that they kind of compiled private dossiers, what we can call them, archives is maybe a little too grandiose a word for these. Unfortunately we have no foolproof way of knowing the extent to which these dossiers came together collectively into an archive because again these are unprovidence texts, these were just pulled from the ground. Nor can we easily establish the boundaries of this archive between himself and his children knowing in particular that his son Buzazu had a house of his own as well in another location. So there are a number of questions about how these texts came to be at this site and if there are other texts at other sites that could give us more details into this trade colony but from what we can see they're copying these tablets and they're sending copies of these tablets back to the capital and keeping a copy locally as well. Okay, so that's one major family at the site and the second is that of M.D. Ilum. So he was clearly a wealthy creditor and kind of the cash flow behind the network. From his family tree we have he and his daughter who are present in very close quarters, what's being called a neighborhood or a block at least by the archeologist and while she was living here he was married to an Assyrian initially who then died and she remarried to an Anatolian and named their children a combination of Assyrian and Anatolian names. So that starts to give you an inkling of the problems we're up against when we're talking about the culture that exists in the Middle East or the culture that exists at this site as evidence from the text that it's convoluted. But a number of these, so these are just like a kind of initial gist of what we can do with some of these family trees and trying to locate them on a plan that's been georectified. So all of this essentially goes into a place where we can hypothesize where some of these people may have lived as well based on as they're, within the text they're talking about our neighbor so and so and we'll be able to kind of provide some geospatial metadata for each of these nodes in the network. But while at the same time so it allows us to further establish like more extensive family trees as well as this starts to become relational data that we see together. And with this we can start to get a sense of a chronological time span. So in demography they call these you know cohorts and they're able to date the ages of people or at least how long they were active in the network based on these cohort time lifespans. So throughout the process the social network which was generated from the text themselves serves as a relational control for each name recorded on each tablet corresponding to the six generations of families who lived at Connash and between Connash and Oshwer between 1950 and 1850 with a resurgence around 1750. So most of this that we see here in the network is really just 100 years, 80 to be more precise but then there's a smaller group of Anatolians that continue after that. Such tools allow us to analyze both the familial and the extended relationship stepping outside the confines of just a family archive which is really how the field so far has been focused into a more full extent of neighboring cliques and partnerships that extend not just within Connash but across the entire Anatolian plain. So what's important to reiterate here is that this network is not a simple aggregation. It's not just text as data being forced together but rather a detailed database that's internally cross-referenced to a number of important attributes. Names of the people being a major one, any place names as well and then a lot of what's taken from the tablets are contextual markers. You'll notice here that there's a cylinder seal. Cylinder seals are a valuable tool for presuppographical studies. And so each of these like components, not just the names but a lot of what's happening on the tablet as well as what the tablet itself is made of. So we're getting, we've been using a PRX scanner, PXRF scanner, sorry, for some of the tablets in Ankara to establish perhaps to assess whether we can distinguish between clay types coming from local Anatolian clay versus the Syrian clay a thousand kilometers distant. Now of course there are other clay sources in Anatolia as we can see here. So based on the knowledge of the sites that these Assyrians are mentioning in their texts, we can show the range of villages or cities if you will, colonies at least in outposts within which the most prominent merchants were connected. So Pushu Ken as we mentioned before by far the most influential but looking at his influence within his own neighborhood as well as across the complete network, we're able to see his central position in the network had to do with his ability to forge more numerous relationships on the boundaries of the network. Kind of a cosmomedici of his time in that his ability to have this high position in the network was his ability to make relationships in contracts with people that extended outside of the regular Syrian network and gave him access to local networks. So that included wealthy investors at the capital washer but as well as the people on the ground who had access to copper mines in Dorhumit up north most likely, I mean they're within some range. And so using the computational approaches the networks allow us, we can go through this process of geo rectifying individuals and their attestation of these place names and this can help fill in the gaps of what we don't know where certain places might be based on the push and pull between the social network and the locations that they're mentioning. But of course even with working with a data driven analysis we're always making assumptions and interpretations along with our new discoveries and working with network models. I've learned how to adopt this structural scaffolding in order to accommodate this kind of like fuzzy nature that we're dealing with when you don't have control on the ground using probabilistic distributions with as much granularity and transparency as the texts themselves reflect. So again each of these nodes represents a tablet or an entry in a tablet and all of that is recorded at the level of the artifact not divorced from it. So then we're able to show the density of the entire network atomized over the course of the documentation and in doing so we can see the degree to which particular merchants traveled across the 40 mentioned place names in the texts. This is more of an affiliation or like which places were mentioned the most in terms of the heat map on the background not necessarily with like certainty as to where these places are located. So this is ongoing research of course with archeologists that continue to return to these sites and in a sense to try to find out where especially Prushadum is it's a still very kind of debatable place. But that matters less from the texts as far as the texts are concerned there are a certain specific number of times that place has been mentioned and we can then get a sense of who's going where when as well as which places appear to be the hubs of the network in terms of the trade that's occurring there for different reasons of course. Bronze refineries here and here, copper mines, Kanishk itself was kind of the portal to a lot of this what's called this copper circuit at the time. And then a Zalpa probably is like an entry pole offshore here then and with a lot of the other materials coming from presumably Iran and Afghanistan for the tin and mostly that tin is what's being carried on donkey back over the thousand kilometers to be traded for silver in order to alloy with copper for the bronze age. So this can be done then on a large scale as well as a local scale for establishing more proximate provenance. But by fixing these texts that we know came from certain households we can then probabilistically aggregate those texts with unknown provenance through the social component. And other mentions of houses and offices and contracts for example are helpful in that regard. Using the ties that bind them socially essentially to the neighborhoods from which they came. The results of this work have sparked a new interest in the utility of these ancient texts especially from archeologists who've largely ignored the texts because of the initial kind of divorce that happens when texts are found in excavation hurried and sent off to a museum to be numbered and curated rather than contextualized on the site. So I've been working more with Turkish archeologists on ways to correlate not just the family tree data but these kind of neighborhoods that are taking shape textually with the private houses and the human remains found in the houses. This includes they have a lot of teeth samples from humans from different rooms and working with an archeologist who's looking at strontium levels and her theory is that she can see a division between Anatolian and Syrians in the strontium levels. Well if that's the case then I can explain based on who may have been living in the house that she's looking at whether their names were a Syrian or Anatolian. Not actually anything that the strontium level would dictate necessarily because as you saw in some of these family trees they were intermarried. And so it's an interesting question at least though and one that peaks the curiosity of archeologists to say but can a data set from the texts inform a method that we're using right now to try to establish who lived in the house. Okay, so lastly it's equally important to me that this process of digitization and online curation or curation through databases attempts to right the wrongs of kind of this past that we're dealt with in the Middle East and Near East rather than ignoring the long colonialist legacy which is often accompanied archeology in this region. Today's digital archeologists can make their data public contextualizing their history rather than incentivizing the purchase price of these artifacts and tablets don't cost much so it's easy to buy a thousand at a time as we've seen in the news recently with Hobby Lobby purchasing 5,000 and 600 and getting caught again for another thousand. So and they're easy to find as well it's just a matter of going in your back yard and digging them up I guess if you live in the right place. So by leaving the artifacts where they're found we can hope for better practices of contextualization through digitization in the future and that's really like where I feel like the work that I'm doing is trying to push that envelope and saying yes there are limitations to working with the bull colossi but if you can digitize them and if you can print them in 3D then it changes the game and even without 3D printing we have so many ways of working with the data that we can kind of begin to remove this need of putting our hands on everything and I think that that's really where a field like mine in a seriology and ancient Near Eastern studies has a chance to shine now because there's no way we're gonna get our hands on these in any time soon. So thank you very much for your patience and interest. I'm happy to take questions but I'm also happy to kind of demo the network software a little bit more. I imagine that that's maybe where interests lie less so in like my Google Earth renditions of these houses more so in the network analysis and maybe the statistical components of that because that's been highly useful in this work using the statistical measurements of the network to determine who's who in the network and then why they're there. But I'm also happy to take questions. Okay, cool, perfect. Okay, let me mirror my screen, that'll be easier. Oh, did it do it? Show mirroring options in menu bar? No, mirror displays, there we go. Okay, and I think it's probably, what if I optimize for mine? Yeah, that's better. Okay, I didn't tell it to close. Shoot, okay, I'll reopen it. Geffy, I'm using Geffy as kind of a starting point and from once you get data into Geffy you can export it in JSON files and graph ML files into more sophisticated network software like Cytoscape. But I recommend it and I teach it in the classes I'm working with this semester and last year too. It's open source and it works on any type of computer. Let's see. So I have four different networks here. I'll start actually with one that we've been working on and unfortunately the visuals are not that great. So within Geffy you have this nice front end visualization tool that's useful if you're getting down into the details and you wanna know exactly what one of these nodes is so you click on it and then it gives you the data that you've encoded into the model. But if where you live on a day-to-day basis it's usually in the data table itself and you're giving attributes to the texts that you're working with let's say. So each of these are individuals mentioned in texts. We have, we've signed an ID to a particular person and we note in this case the two numbers that correspond to the texts. Now this is work that Nick Feldhaus and I have been working on this semester with a URAP team. So all this data is coming from ORAC, the online corpus of Kineiform literature and texts. So we're starting off with texts that have been lemmatized which allows us to do something really cool which is for each individual we can say what their role is in the text in really like simplistic terms. These texts were three texts. So from day-to-day from 2100 to 2004 BC about 50 to 100 years before the oldest Syrian texts or the oldest Syrian texts were written. And so there's really just like a 50-year gap between these two which made it a really interesting dataset for me to work on subsequently. So within, so aside from what we can pull from the texts themselves including the role and if there's a profession that's mentioned, that you know the different texts that these people are mentioned in and then the years that are joined when we join texts together. You notice that there are a couple of other statistical measurements that were generated called authority, hub and eigenvector centrality, the number of others. These are like eigenvalues. Eigenvalues are essentially the yin and the yang of the network. So if you think about a social network you have people that you point to like say let's say people that you tend to write emails to and you have people who tend to write emails to you. So there's a directionality that occurs within every network. And this directionality is really valuable for telling us who's who in the network. So if I'm the type of person who only rarely writes an email but I receive emails constantly you know then I would be someone who's considered in the know, is equivalent to the hub. So and then the inverse of that would be an authority. Someone writing and telling people do this and that. Eigenvector centrality has to do with the quote unquote leader. So it's a leader metric. It's someone who doesn't necessarily have a lot of ties or relationships to other people. Maybe fewer relationships to others but those relationships that they do have are with people who have very many relationships to others. If you think of like the CEO around a table he talks to maybe seven people but those seven people are talking each to like hundreds of people probably. So that CEO would have a very high Eigenvector. So these are statistics that can be measured very easily within this setting. Once you've given the properties of the network based on these types of the yin and the yang, the pull, the push in the pull, who's receiving something? Who is sending something? Well who is the source of something being sent? All of that was documented on these texts. And just those three categories who's receiving, who's sending and who's the source will tell us immediately a lot about the statistics of the network. So you can come in and run these. It's just a run button. Average clustering coefficient, Eigenvector centrality. Which of course is dangerous for someone who doesn't actually know what these algorithms do but the nice thing is is that it's contextualized right here in front of you and then you can come to this place in the appearance and look at, let's say in the appearance tab we'll look at the size. Let's make the size based on the rank of Eigenvector centrality. With the minimum being 10 and the maximum, let's make it 50. And then the network changes its property and you can begin to see. So let's do that same thing for color now and choose the same measurement. Let's see, maybe if I turn it to white it would be more visible. And we can then see, okay, so it's not, so it's Shulgi Irimu, Nalu. I'll admit to some degree but he's really the leader of this. Now let's take a look at this guy. Shulgi Irimu and let's look at, okay, so we don't have any profession. We don't have any profession listed on the text. So if you're just close reading these texts, you wouldn't know just innately who he was. He doesn't declare who he is but if you look at his role, he's every time recipient in every instance and all the texts that were joined together he's the recipient of the goods being traded here in this colony. So that tells you something. Eigenvector centrality is equivalent to recipient. Okay, I mean this is like really basic and of course this all has to be contextualized more but let's do it again. Let's look then at go back to size and change it to the hub and apply that. And all of a sudden it changes properties entirely, okay. So let's look here and see. This guy and we have to probably change the text size too. Okay and now it looks like a totally different network and we see Abashaga. Abashaga's profession again, he doesn't list anything but the role that he occupies again almost every time except with a few recipients, he's the source. And these things are mostly animals. Draem texts, this is the Draem archive which are mostly like animals being traded in a peripheral place on the Sumerian kind of or three administrative state at the time, city state. So the nice thing is that you can then go in and textually understand why he's the hub. What is this statistic actually telling me but contextualizing it by the texts? That's really helpful because otherwise how do you teach eigenvector centrality? Well you do it through linear algebra but who's gonna take class on linear algebra to learn what eigenvector centrality is? I mean I don't see any hands popping up. So instead we can teach it through a more hands-on place that actually gets them rooted in what's happening with context that they're already interested in. So let's change it one last time to this authority metric and we see again the network changes entirely. It's now we're looking at Aradmu with Shugi Irimu who is the leader being like a prominent node but not nowhere near as much as Aradmu. And here again, no profession. So these are people who are just assumed, they're just assumed that they're leaders because they don't have to put their title down. And the texts are pretty clear that they have these places and then we look and in almost every instance he's the representative, he's the via, via. So something's taken by this guy to from the source to the recipient and he's the representative. Now it's pretty cool to have this kind of control over 15,000 Sumerian tablets. Like this is the aggregation of the tablets and then to be able to jump into the back and say, okay, well, which texts are we talking about here? And then to like click on it and drop this into the online database. I mean, it's immediate and you can be working with the texts at the same time very easily. And all of this is open source and it works just by importing spreadsheets. So there's not a huge hurdle to enter into this realm if you dare, right? Because it's always about the trick of it isn't necessarily what we're looking at now. This is what's called the nodes list. This is the data that each node carries with it and the metadata. The trick of it of course is the edges. Network analysis measures relationships, not just data points like most databases but the relationship between data points. And so we're going, we're assign each node an ID and that ID is used as either a source or a target. And that's the explicit relationship. It's directed and then that edge is given an ID and it's given a weight. That's the important part. That gives it the kind of structural push and pull in the network. And then here we also provide the years that are tested for this edge and the text in some cases. So that's one example. And I think like a real prime example, one that we worked on recently, there's of course others. This is the old Assyrian network. It's still being disambiguated by hand because there's so much proponomy within this text. So there's so many people using the same name over the six generations and that we have a grandfather who will name a grandchild his name before he dies or maybe as he dies, the parents name that. It mostly happens through the male line but there are a few examples of females named after their grandfathers and then you get proponomy happening across genders. The cool thing when running the statistics here in comparison is that instead of the source and the authority and the hub being three separate people, they collapse. Two of them collapse under Pushuken and Imdi Ilum. These two people that I talked about earlier in their family trees, one being the wealthy creditor, the other one being the main center of trade. So it gives us a way now of like comparing networks that we've never dared ever do before. Comparing Sumerian were three administrative network with this long distance trade network in Oshur. Well, that's risky business but at least in this way we can do it from a kind of sound methodology that's been laid forth while still questioning assumptions and interpretations all the while but it's a structural scaffolding that allows us to kind of see, can we follow this? Can we equate an apple and an orange? What properties do they have similar? Well, their statistic properties are kind of, it's there either they're connected or they're not connected. That's binary. I mean, you can apply a weight to it but this gives us a way to compare something, the context of which is huge, 22,000 nodes versus Draeum's 1,000. Then jumping into, say, Ebbla and the Ebbla archive that has been reconstructed and the different shapes that this is taking has to do with the text typologies. This is focused more on the texts themselves, textiles and pots and pans and utensils and the different, so there's kind of this like a leveling effect that takes place because you do have to atomize it all but because it all takes place within the nodes list and edge list in the background, there's no distance between you and your data at any point. Like you get to be the one who's curating it. On the back end and playing with it and coloring it on the front end. So that's really, I think, a helpful tool, especially since a lot of archeology is done in the privacy of your own office. And yes, these files are easy to share as well and that's the nice thing. Like this isn't my work. This is coming from Italian or a seriologist, Massimo Maoki who works with the database and is interested in analyzing it for different kind of structural properties but is more of a text person. So in the end, like that's really, I think what this and other tools like network analysis and text analysis can do is start to build bridges between archeologists who work with only with bones and strontium values and philologists who only work with tablets and neither the twain shall ever meet at the same conference even. Unless you can find some relevant way of bringing these two together and then all of a sudden both of your data sets have taken on an additional dimension. We're adding new dimensionality to this. So that's kind of the argument I make for these types of tools. And I think that if you're looking for more dimensions to your work, it's a great way to go about it in a transparent way. One that you're not ending up with like some black box results at the end but rather something you can control to the nth degree. Yeah, sure. My pleasure. Yeah, I'm happy to. Yeah, it's called Gefi. You can see it right there, G-E-P-H-I. It's a dot org. And so it works on all the different types of computers. Yeah, yeah, sure. Yeah. Yeah. And yet you had individuals going back to individual rooms. Is that because of what you learned from doing your textual connectivity analysis that you could actually pinpoint room blocks or something like that? Do you work backwards? I work backwards. It's working backwards to the archeology. So there were a couple of, so the early archeologists who came to the site started it in the place that was looted. So they saw, ah, this is clearly where the tablets have been coming out of the ground. Why don't we just start here? And so that's how, in the very initial report in 1948, it's one of the best actually that they've ever done, unfortunately. Or fortunately, I mean, you know. And there they were very clear about like what few additional texts can they, then from that first season they pulled. I didn't call it an lawsuit. That's right. And then it's clear this is the same household because of the continuation of an event like a lawsuit that's unfolding. And then we'll have additional, like records or contracts from that lawsuit that correspond to the thousand or maybe fewer that we have on provenance. And then circular way back. Yeah. And yeah, it's just like backwards engineering. Can we find it this way? In a lot of cases you can. And like that's one of the problems is that for the old Assyrian stuff, there's a lot floating out here, out around here that is not connected at all to anything. And that will be, that's a continual project, but the fact remains of the 30,000 texts that have now come from this site, there's more that they could find each season. Like they're not done with pulling tablets out of the ground. So it could very well be that this chunk over here has just not been excavated yet. And though the names that are in there are unique to us now, but they'll become known to us as they continue to excavate. Or you know, they are unique names in which case like you can't do anything with that anyway because so Moans Larsen who edited one of these archives that was excavated in situ and with proper provenance and loci points. Turns out that within that archive, that local archive that his family used, there were other people who weren't related to the family that were depositing texts in there. And those people would sometimes put texts in without any names associated whatsoever. They were just commodities and receipts. Okay, well what do you do with the receipt that could be from somewhere else? Well, you do PXRF and you say where did the tablet come from maybe? Or you know, but you're limited at some point. So there's always gonna be a number of tablets that you're just gonna have to say, we think it came from here but we don't know where. And we definitely won't go to putting it like as a point on the ground with geolocation. But then in these kinds of clusters and maybe typologies of tablets that we could then work with, there's like at least 40 types of tablets that they wrote here, 40 different types. And those are ranging from legal briefs, very formal legal briefs and business contracts all the way to personal notes and memos and rits of divorce and things like that. So it has a full range of the types of texts we could then say, okay, well, we've never found a writ of divorce in this place before but they all seem to be coming from over here at this like office where marriages and divorces take place. Yeah. The display, is it synchronic? Everything is slurred together. There it is. It moved through time because we didn't have a control rate anymore. That's right, yeah, it's a really good point. So within here, and especially for the dram texts, what we're working on, so we just started doing this network and the nice thing about the texts from DRAM, these 15,000 texts that we have from DRAM in Southern Mesopotamia Sumerian tablets is that each one has a date on it. So every tablet is dated. And sometimes to the day and the month and the week and the year, so you get all four and one and you're like perfect. I know just when it happened. So there is a way to do time sequence and there are simple ways and there are complex ways. The one of them will be like kind of a Boolean form in that you say like based on which year it displays certain parts of the network. Now it's always like incredibly easy to just create a new network, copy to new workspace. And there I've just sent this subset that I just randomly selected to a new place and now it's a subset of the network. So it's very easy from the huge kind of starting from everything together to then just piece off. No, I just want a year of it. Let me just take this year, let me control for that because maybe I want plus or minus five days and then I can make a network that way as well. So there's a lot of variability in terms of like what you see if that's like how you're working with the network in the visual here. Or you can just simply work with it in the data aspect of it behind the scenes and make your subsets there. But the nice thing is that there are a lot of plugins for Gefi and these are constantly being updated because it's open source software. So there's a lot of development going on for this tool and you can go through and see that they have a good documentation for each of these plugins here that allows you to look at scale parts of the graph with a scale factor, minimum spanning tree, calculate minimum spanning trees. You have, let's see, the ones that I think were from time I already installed. Let me see here. Yeah, I mean a lot of the, and they're all documented except for the standard ones. There's some really good descriptions about what these do. So the time span one, I'm forgetting the name of it, but it basically gives you this interval metric here. You see timestamp here and there's a timestamp and then there's an interval one so that you can give it a range and then it will display based on that interval if when you like do a filter. I mean, that's the other nice thing about this is that it has this really great, like you can filter on attributes or any kind of like dynamic variable. I'm gonna start clicking things and it's gonna crash because it is open source software. It's always a problem with demoing anything, but you can see that you can then like create an EU network very easily by just dropping that component in here and saying what the parameters of that network are within the depth of one, two, three or max. And you can do that with some K-core or an attribute. So let's say a year. So if you have within your dataset a column or an attribute called year, then you can filter very quickly based on that and export that filter to a new workspace. So it's really adaptive and interactive in that regard. So yeah, you're right. Like at the time it's kind of like you see it and it's all like, well, when is this taking place in time? Yeah, end space for that matter. There is this geo layout that allows you to graph it based on geo coordinates that you've put it in X and Y values. And it'll read those in as well if you've labeled your headers like X and Y or lat long. So they work really nicely in that regard with what I need for archeology. But yeah, it's like kind of getting the, getting filtering down to what you need is like, you know, kind of always the challenge for everything in my fields because you have so much data. Well, thanks again.