 Welcome, everyone, to the Understanding the Language of the Digital Humanities Lightning Talk. I'm Rebecca Cummings. I'm the Research Data Management Librarian here at the Marriott Library. So I'm very excited to have this many people here. I actually really had no idea. This is our first event we've had in the Digital Matters Lab space, which is technically more that way, but this had a little bit more room for this kind of presentation. So thank you, everyone, for coming. So if you haven't done this type of event before, the way that it's going to work is that each of our presenters get five minutes to just give a very general overview, a non-technical overview of a digital humanities topic, and then a couple of our presenters were kind enough to take two topics so they get ten minutes. And if any of our presenters have any problems with the technical aspects of this, please let me know because I know everyone's computers work a little different. So this is an overview of the topics that we're going to look at today. First we have David Rowe, who's going to give us a general overview of what digital humanities is. Brian McBride is going to talk about APIs and open APIs. Lisa Swanstrom was one of the people kind enough to take two topics. She'll be talking about text mining and text encoding. Nina Fang will talk about gamification. Elizabeth Calloway will talk about topic modeling. Leona Yifan-Sue will talk about sentiment analysis and machine learning. Anna Nitrar will talk about metadata. Jeremy Minty will talk about linked open data, and I'm going to be our last presenter today talking about data curation. So without further ado, I'm going to turn it over to David Rowe, and I'll even help you get your slides cued up. Oh, and if you do have questions, let's actually save those for the end, just to make sure we can actually get through all nine presenters. Good afternoon. Thanks, Rebecca, for setting this all up. She's been working tirelessly to make sure that we get digital humanities up and running at the year. So thank you very much. I also want to start by reminding everybody that we have a symposium in February, the Digital Humanities Utah 2 symposium. Our keynote speaker will be Alan Liu from the University of California, Santa Barbara. And we also have a special guest coming from the National Endowment for the Humanities Office of Digital Humanities who will be here to host a workshop and possibly have some one-on-one time with you if you are interested in getting funding for your digital humanities projects is really a good opportunity. So if you have an abstract that you'd like to submit, or if you want to participate, please make sure to register. OK, so let's start. What is or are the digital humanities? It's really a broad umbrella term. I'm going to give you, I'm going to just add to your confusion by giving you my definition and sort of a broad overview of some of the sub-disciplines. But it's kind of surprisingly long history. The first thing I want to talk about is that digital humanities is kind of being shaped by a debate. It's a very healthy debate that's happening right now. We're not even sure how to define digital humanities, but through this kind of discussion we are coming to an understanding of it. But there's roughly two camps. There's the theory camp and then there's the praxis camp. The theory camp sees digital humanities, I'm sorry, the praxis camp sees digital humanities as a means of reinvigorating the humanities. They see software platforms and technologies as a means of creating new perspectives, new critical terrains. These are the kind of people who are putting out those sexy visualizations that you see, getting a lot of press in the Atlantic, the New York Times, all that large data corpora being mined for information. The theory camp, though, they may well be invested in digital humanities tools, but they want to put theory first. They want to put critical theory first, right? And they're afraid that the emphasis on tools and building might actually allow some of the larger, more important questions. They see digital humanities as a field that should not be dependent on technology, but that the logic of technology should inform and guide new areas of humanistic research. So this kind of manifests in this binary. Between building and theory, between method and methodology, hack versus yak is the colloquial term. But there's nothing really inherently wrong in building tools, I don't think, but the tradition in which digital humanities is steeped in, they can be kind of accused of being uncritical. There's a long tradition of humanities computing, for example, and that's mostly affiliated with the fields of linguistics, for example, who have used computing for a long time as instruments, but not really as objects of inquiry themselves. So the danger lies in our rush to cutting edginess that the larger, more important questions will be neglected or ignored. But I think there's a room for a middle ground that the humanities needs to adjust its paradigms and embrace a more constructivist model that is just can't simply critique, but it must critique by doing. So there's really quite a wide range of subfields in digital humanities, some of which have only recently coalesced, and I'm sure there's more to come. All right, so platform study, software studies, digital forensics, these are the studies of the GUI interfaces, the graphical user interfaces, the interfaces that you use every day, the Windows interfaces that you use on your computers, the mechanics of word processors, how do those kind of digital environments affect the processing and production and consumption of knowledge? Tactical media, media ecology, this field sees media as participatory. They see technology as sort of the conduit through which different factions can use, take advantage of the network, distribute in nature of the internet to, for example, mobilize politically. Encoding data mining, these are the kinds of works in which you have XML tagging, for example, metadata studies, the architecture of databases and archives, topic modeling, text mining, these are the kinds of works in which you are interested in large corpora information and you can find new perspectives by using machine learning and algorithms. History of the Book and Media Archaeology, this reads print as media. Print is a technology, right? And placing things like the e-reader in the long history of the book or studying how does reading change when we go from the print book to the screen? Do we read in a more non-linear fashion? Do we read in more, do we skim more? So at its core, I think digital humanities is really a perspective on the humanities, right? It's informed and facilitated by technology but not really determined or dependent on it. It's useful for reexamining old axioms for uncovering new terrains of humanistic inquiry. And I think that's really when digital humanities is most interesting when it kind of facilitates the boots on the ground humanistic research that we are all very familiar with by uncovering new areas of research or filling in the scholarly gaps. And the interdisciplinary and collaborative nature of digital humanities ensures a lot of fresh new insights. Thank you. So Rebecca asked me to speak about APIs. I know it's a very exciting topic for everyone. It's exciting for us actually because it sort of gives us a lot of tools that we need to perform a lot of things we do today. I noticed there's a few tablets in here, there's a few laptops. And those are all connected to systems via API calls and everything like that. So I wanted to give a pretty short introduction to APIs very overarching view, nothing too technical. And here's an example that I think everybody has probably heard this is a set of routines, protocols or tools for building software applications. And it's sort of a neat way to think about it is that if you take a second look at the person in the middle, a good way to describe an API sort of as a middleware or middleman because you want one system to talk with another system and the systems might be using different architecture they might be using different protocols, gotcha. So what you need to do is you want these systems to talk to each other but oftentimes they're owned by different companies or anything like that so you have to develop APIs so they can communicate to each other. So I think that's a pretty good way to explain it in the abstract. Rebecca mentioned open APIs. Open APIs are typically APIs that are open to the general public to use. I'm sure some of you probably know that Google has APIs available for their cloud, their maps and a host of other platforms available to them. But there's also closed APIs that are restricted to vendors so you have to have a license to use the system. You can't interact with the data or anything like that. So typically we use open APIs in our development here at the library. What does an API look like? Take this with a grain of salt. So essentially this is just a simple curl request and it's just sending out a request to this controller here and this is the response you're receiving if you notice that there's a very specific sequence of events and this is sort of what is defined in the API. So when you make anybody who makes this request is gonna see this type of response so you can programmatically adjust your program to taking this data from the API. Here's another thing. This is just one example. This is a JavaScript object rotation and this is just a response. And you can look here, you can see the names. You have the data, the type, the identifier, the attributes and these are all defined within an API. And within those, the API will actually describe the format or the standard in which the data will be sent back to your application. Some people like to use different standards for timestamps but what the API will do is actually tell you what method it's using, what ISO standard or if you're not using an ISO standard and it will put the data in a format that you know how to expect. And so what that does is it allows you to build your applications in a way that can integrate with the other systems. So here's the thing. Has anybody used a cell phone today? Raise your hand if you've used a cell phone. So it's pretty much everyone in the room. Does anybody have a Nest or an equivalent ECOB or anything else like that at home? Does anybody have an Alexa or Echo at their home? These are pretty cool. Wi-Fi enabled lights where you can actually change the color saturation that's being emitted. And I think too, especially considered that we bought, we bought from my house, a RATIO and this is an integrated irrigation controller that actually uses APIs to contact application on your cell phone and you can control it wherever you are as long as you have a network connection. And the other cool thing is that it actually uses an API open standard so if there's a public weather station in your area you just tell RATIO your zip code. And what it will do is actually go out and look on a central database and see if there's a public weather station. If there's one you can associate with your irrigation controller. So if there's anything like a weather pattern change, say it's cloudy, okay, I don't need to water my lawn per se. Or if it's raining, I don't want my lawn to be water because that's just a waste of water. So these are all sort of mechanisms that we're using APIs to call to different systems. So it's just this layer of integration across the platform. Do you use Facebook on your cell phone? I mean, I have to raise my hand, I'm pretty guilty of that. But it's using APIs to call to Facebook, to interact with everything. We can go over the graph API if anybody wants to later in the day. So the other question, anybody interested in using API? I just want to dabble with API. It's here, there's a couple of services out there. They're sort of the if then statement. So what you can do is you can use these providers and you can design recipes and they're really, really basic commands. So what you can do is you integrate different services. You can integrate, say if I get an email from Gmail, you can integrate this into Gmail and say if there's a document attached to it, I really don't like Google Drive, but I like Dropbox. What you can do is that this will monitor Gmail. You can set the recipe and then it will do is it can take that actual document that's attached in Gmail and sync it up to your Dropbox. So I mean, that's sort of a good example of Dropbox doesn't necessarily necessarily talk to Google. But through this API, you can actually have that functionality that you're looking for. And here's another vendor. So sort of wrapping up some notable interests in APIs for the digital humanities realm. We heard about machine learning. That's actually something that Google has pushed out rather recently that looks rather compelling that we've actually started looking at. Google Maps. I think we've got a couple of collections here in the library that we actually have coordinates related to the collection. And we do Google Map overlays. CloudVision, this is a new system where you can actually look, Google has special software that can actually take apart the images and look on at a different perspective. The speech recognition translation. So what this really does is say if you don't know a certain language, but you need the documents translated, you can actually use these services with writing some new basic scripts to connect to these APIs and have the translation services done. So it's another way to get a massive load of data programmatically into these systems and have it come out in a format that's useful to you. So the other thing too is cloud services. That's a big thing because the one thing you can also think about APIs is maybe Microsoft Azure, maybe they use Microsoft platform, the server platform. Maybe Google uses their own home ground system. Does that really matter to developers? Does that matter to managers? If what backend system they're using as long as they can interact with that system? And that's the thing that API is like, I need the compute, I need the data, let's, I don't care what the backend is, let's do this to the business proposition. So if that hopefully answered some of the questions that may have existed beforehand or not, but if anybody has questions, shoot me an email or if there's anything else. Or at the end, you'll get a chance. Yeah. All right, thanks, Ryan. Thank you for putting this together. I think this is such a great idea. So I'll just jump right in. My first term is text mining and what I thought I would do is give you a definition of data mining in general because text mining is kind of a subset, you might think, of data mining. So data mining, a great kind of general all-purpose definition of data mining, the non-trivial extraction of implicit, previously unknown, and potentially useful information from data by means of a digital algorithm that sweeps through this information and hopefully plucks out meaningful patterns. When we think about data mining, we often think about, in general usage, the data mining of consumer habits, for example. So if you are a listener of Pandora, or if you are a reader on Amazon, or if you watch Netflix, you've probably seen the result of the data mining of consumer habits. And any time, for example, someone says, ah, this is a song that you might like and you actually play it and you actually like it, here is the kind of uncanny consequence of consumer mining data consumer habits. Text mining is quite different from data mining in this general sense, because in text mining, particularly my area, which is the text mining of literary texts, right? Artistic, aesthetic texts. It's a very different beast, and the difference hinges on the distinction between what we might call structured and unstructured data. So every time you listen to a song or purchase a book or watch a movie, you're generating information about your consumption habits that are really easily trackable. You might think of a database, for example, or a spreadsheet where you have the customer's name, your location, the name of the film or the song that you liked, when you watched it, and all of that's really well organized. You could think of just specific individual fields for all of that information. And so it's pretty, I'm not going to say easy, but you can see logically how you would be able to compare that very precisely organized information against the habits of thousands, thousands, perhaps millions of other users. Literary texts are structured in an entirely different way, and we call that type of text unstructured because it's not structured according to a database format. There aren't fields in the way that we write literature or that we're having a conversation right now. It's not to say that there isn't a different type of syntax or structure. Of course there is grammatical rules, syntax, all of these things, but they operate very differently from the way that a database of consumer habits, for example, might operate. So text mining is looking for those syntactical patterns in unstructured data. And you can have data mining on really small scales. This is a project that I've been working on. I've been trying to come up with a few tools for actually close reading literary text through a very simple type of data mining or data reading, where I wrote a very simple program that would go through a text, count up all the colors that appeared in text and then spit out a color palette based on the percentages of colors that appeared in the text. So here, nothing too surprising. This is Edgar Allen Poe's Mask of the Red Death. So red appears very frequently in this text, no surprises there, but I have found some really interesting examples of this type of palette usage in texts that have actually provoked further questions. But I thought I'd share with you a much larger, grander scale example of text mining. This is from a project by Dan Cohen. I don't know him, I'm just familiar with this project. But he wanted to, as a Victorianist, question some of the long-held assumptions about Victorian culture and letters. And so he actually worked in cooperation with Google Books using their Ngram technology. And he worked with over 1.6 million texts. That is real kind of text mining, right? With my example, it's so micro that it is a little bit embarrassing. But this is large-scale, very useful, very provocative approach to mining literary text. And what he did was look at the titles of texts during the long 19th century in order to question in particular a long-held assumption about the crisis of faith that was said to be happening during this time period that because of introduction of texts, for example, like on the origin of species or the descent of man or concepts such as entropy, that there was a distinct traceable falling off in terms of religious faith. What he found, however, by going through 1.6 million titles of texts is that there was a sharp decline, but it happened much later than people had thought that it would have happened. It happened peaking in 1850 and then dropping off until World War I. And what he suggested is that going through this text mining of 1.6 million texts is not providing an answer to this question, but it is very interesting because what this points to is that there was at least a drop off in the publication of texts that use this religious language in their titles and that this warrants further examination. So what I like about this project is that it's not offering solutions, but it's kind of provoking questions for further analysis. So those are a couple, tiny scale, large scale text mining. My second term, excuse me, is TEI or the text encoding initiative. And if you think of text mining as going through unstructured data, trying to look for patterns, you might think of the text encoding initiative or text encoding in general as a way of salting the minds, right? Of providing structure to unstructured information so that you can more efficiently look at it and see, look for those patterns. So I don't have a ton of experience with text encoding, but I was fortunate enough to work with the Woman's Writers Project with Julia Flanders where she taught me kind of the basics of text encoding. And the text encoding initiative was something that was started in the 1980s. And basically people have been encoding or adding structure to this unstructured information for a very long time. And what different groups realized is that we could all benefit from having a shared set of standards about how we decide to encode this information. So to do that, I'll give you just a short example. Here you have a wonderful tried and true poem, very canonical, and is also one of the most frequently cited in terms of text encoding, I think because it's very short. But here you have William Blake's The Sick Rose. And if you were to add encoding to it, if you were to salt the minds in terms of the text encoding initiative, what you would do is go through this poem and give some information about it. So you would then be able to extract it later more efficiently. So for example, if you were making an anthology, you would add a tag, anthology at the beginning, and then you would have another tag, poem. And then he would have a tag, here's the title of the poem, The Sick Rose. And then this information stanza tells us that a new stanza is starting. Each line indicates a new line, and at the end you have the backslash, which indicates the end of that line. And so on and so forth until you get to the end of the poem. And because you're making an anthology, put a little note, more poems go here, and you have the end of your anthology. And the idea here is that your reader is not going to see this information, but if you are in fact making an anthology, this makes it so much more useful. I would say that I wanna look at poems from the 17 or 1800s. I have a specific date in mind. If I've encoded that date, then that means I can just search very quickly and get all of that information that has been tagged with that date at my fingertips. And so I haven't actually worked with any of the extraction end of text encoding, but I find it to be a really useful tool, even if it is a kind of example of salting the minds. Thank you. Okay, so not using digital technology might seem kind of ironic in this talk, but I think it's important to emphasize the use of non-electronic mediums, alongside digital mediums as well, because they operate in a context, and it points to that kind of context. And I think it's especially useful to think about that in terms of pedagogy. Also, I like whiteboards, so. So I'll be talking about gamification, which is the use of game elements and non-game context. And again, we have this word context, right? And so I think it's really important to think about gamification as how do we assimilate or incorporate game elements into our classrooms in a way that's really effective and useful for our students. So using context is a way to highlight gamification. The game elements that we're really talking about are ones that come from video games. So things such as points, leaderboards. And points are kind of self-explanatory. They're sort of a reward system, so students can receive points maybe for accomplishing a task, like just watching a video or finishing an assignment. Leaderboards are a public way to show rankings of these players and their points as they progress. And badges, so after completing a task, you may receive, you may unlock an achievement and receive a badge, such as a thesis wizard or something like that. Also narrative, which is the use of story elements to put a task in context. So you've probably seen the use of gamification in marketing with things like the monopoly game at McDonald's, right? Where people are urged to buy food, to get pieces of property, which they can accumulate towards rewards. And we've also seen game elements used in an education, which Quest to Learn is the first charter school that's really based on gamification. So students will pursue quests, such as missions to solve a math code. Levels replace grades, so you may see very, instead of, levels may replace grades. And then there's also boss levels, which are two-week intensive units that equate to a midterm or a final. And students can choose roles within a group to tackle these kinds of tasks. This is a kind of deep gamification. And we can think of gamification as shallow or deep. And shallow gamification we can think of as structural. So it's the application of game elements to propel a learner through content with no alteration or changes to the content itself. The content does not become game-like, only the structure around the content. So an example would be a learner gaining points within a course for watching a video or completing an assignment, but there are no game elements within the assignment itself. The most common elements in shallow gamification are things like points, badges, leaderboards. Deep gamification, on the other hand, is the application of game elements and game thinking to alter content, to make it more game-like. So for example, maybe starting a course with a challenge or a mission, instead of just going through a list of objectives. And there is some research to show that shallow gamification may increase motivation temporarily through a novelty effect or maybe decrease motivation even in the long run. But I don't think it should deter you from trying out these kinds of methods in your classroom because there's just not enough data right now. We haven't seen enough studies on it. And so we don't really know shallow rewards are definitively detrimental to motivation. I do think it's important to place everything in context. So the design of the gamified course is very important. You wanna try to connect your game elements to class content, which is why I'm currently conducting research on the effect of a fictional narrative on a composition class. And so the goal of the student really aligns with their goal as a researcher who is trying to create a report for a alternate universe. So the material they read and study plays directly into their dual role. And I'm hoping the narrative context will increase motivation and engagement. It's kind of a shameless plug about my own research. So we'll see at the end of the study by this semester. Thank you. My mission today is to talk to you without getting too into the weeds as per my instructions about topic modeling, which is a little bit of a difficult mission considering topic modeling is one of the more mathematically complex methods we have available to us in digital humanities. So what I'm not going to do is I'm not going to get into the underlying statistics behind how topic models work. And I'm also not going to pepper this talk, this short talk with caveats about interpreting your topic modeling results, even though I think that is an essential element to the genre of topic model papers. And in general, in digital humanities, kind of critically grappling with the limitations and the blind spots of our methods seems to be part and parcel of what we do. Instead, I want to get you interested in topic modeling. I want to get you excited about it. And then if you end up, if I'm successful and you end up trying topic modeling for yourself, you can get as much as you want into the mathematics behind it and the limitations of it. So topic modeling is one of the ways you can do distance reading. To borrow Franco Moretti's term, you can look at hundreds to thousands to millions of documents at once and search for patterns or interesting moments that you want to look at further. In topic modeling, you do this by having the computer put words that appear in your corpus of documents together. It puts words that tend to co-appear together into a topic. So if you have a topic of words, it means those words have a high probability of appearing together in small chunks of your text. When I do topic modeling on, for example, nature NGOs like the World Wildlife Fund, I have my code here and it spits out these topics here, lists of words that tend to appear together. You can also have the computer visualize them in word clouds for you, which helps you to see how distributed a topic is among its terms, whether it's dominated by the first term in that list or whether, as you see here, the first three terms kind of share dominance of that topic. Here you have phishing and over phishing in the World Wildlife Fund. You get topics like energy, fossil and alternative. We have the contact us topic, which is amazing because it groups all the boilerplate about how you would contact the WWF if you would want to. This collection of economic and natural terms can only be sustainable development. And then here I have a topic that's clearly forests and deforestation. And you can do things like arrange your documents over time and then have, calculate for each document the probability that a topic of interest occurs in that document. So you can see here, it looks like the forest topic in the World Wildlife Fund is declining over time. Now this is when you would want to go and ground truth some of your results, especially if you're looking over long time spans, you wanna make sure that these words are being used in the same way across document types and through time, especially if these results are surprising to you. Here it fits my general expectations that as interest and awareness of climate change ramps up, looking at forests and deforestation as our primary environmental threat is on the decline. So this same method can be used to find moments of interest in your big corpus. I'm interested in the ways that conservation is really has sticky associations with development, development that's not always even or good for those being developed. So here I have my sustainable development topic and I can see that when I calculate the probability of it appearing in each of these documents, it has a high probability of appearing in document 25, which I can get the computer to tell me is the 2014 World Wildlife Fund living planet report. And a particular area of that document that has high probability of this occurring is a discussion of the Varunga National Park in which ecotourism is presented as a solution for a local exploitation of the environment, which if you read on mainly consisted of local women fetching water from within the park boundaries in the gorilla habitat. So here I got, you can see how topic modeling pointed me to a really rich case study about labeling one type of use and exploitation of the environment. Another type of use ecotourism was a non-use despite ecotourism's environmental impacts in terms of carbon and infrastructure for the luxuries tourists might expect. And often the literature on topic modeling really embraces topic modeling as a way to discover interesting moments in your corpus. People have been slower or more hesitant or thoughtful about accepting topic modeling as evidence in and of itself. So I suggest that you try it out on the corpus of documents that you know really well for yourself and get into the literature behind the mathematics and all the caveats and see for yourself how convincing you find it as evidence. So hi, I'm Leona. I'm right now assistant professor at the Department of Communication. So today I'm going to introduce two topics, sentiment analysis and the machine learning from the perspective of a communication scholar. So most of the examples I'll give you is probably from what we do in the field of communication. So one of the questions today we try to answer, I want to provide an answer for you is what is sentiment analysis. I think one of the best ways to describe in one sentence is to discover how other people are feeling. So actually you want to know what people are feeling, how their feelings are. And sentiment analysis is from the perspective of a communication scholar. I will actually argue that it's actually part of a content analysis method. So in a traditional communication methodologies we actually use content analysis very often with the purpose to actually analyze the text in order to find underlying meanings associated with the text. So the difference of sentiment analysis, I think right now is actually the first of things is it focuses mostly on the sentiment side. So we try to find out the sentiments embedded in the text and also it's using those machine-based algorithms. For example, like natural language processing, statistic or like machine learning methods I'm going to talk about today. So if you ask what is actually sentiment actually here refers to, I will actually everything the subjective feelings you can think about also all count as sentiments. For example, like attitudes, evaluations, balances, opinions or emotions, all those different subjective expressions can be counted as sentiments. And that's why sentiment analysis is sometimes being called as opinion mining. So however, as far as I read from those literature is if you focus more on the term of opinion mining you focus mostly likely about the data mining side. So you try to go through terms of data in order to find the patterns about people's opinions. And one of the most common goals of doing sentiment analysis is actually to find a polarity within the context, within your text. And a polarity here can actually refers to those of different things and one of the common ones include like agree or disagree and you like it or you dislike it or you feel positive about it or you feel negative about it or you feel neutral about it. So people with the goal of doing sentiment analysis is try to find those binary of composition of the different emotions embedded. And sentiment analysis I would actually argue is one of the most powerful tools in the age of digital media. And one of the reasons I think that people are now paying more attention to sentiment analysis because of the emergence of those digital media. Because right now because of the digital media right now you have those social media you have social networking sites you right now have terms of opinionated user data. So because of the data is an emerging layer so the social scientists and also computer scientists are actually seeing the needs and also the motives to actually investigate more about how to find out the underlying opinions or expressions about people's feelings. And I will say like from a communication scholars perspective like Facebook posts or comments of news stories, product reviews or blog posts all those are very good materials that we actually study in order to perform those different sentiment analysis with different interest or different research goals. And here I'm showing you one of the examples. It was supposed to be animation but I was using PC so it's probably like I'm using Mac now so it doesn't really transform into that. However the goal of this study is I conducted about a study about studying millions of tweets about nuclear power. So I studied within a period of one and a half years and I want to see how people express their feelings about nuclear power change in the one and a half year period. And the goal is because I want to see how this responds to the eternal events of Fukushima disaster. So after Fukushima disaster takes place does that actually change how people feel about nuclear power? And here you can actually see the six lines here. You can think about it as two different dimensions. The first dimension is about the balance. So how people feel like positive, negative or neutral about the topic. And the other direction is a certainty whether people are certain or not certain about their attitudes being expressed. So you can actually see the changes across time of those different sentiments plus their different certainty about their attitudes. And you see the green spikes here is actually the big spike. This one actually refers to a pessimistic certain tone which means after the Fukushima event which takes place here in March 2011. So after the March 2011 Fukushima disaster happens actually people express lots of very negative perceptions of nuclear power in the very certain tone. So that's one of the examples I show you here is how you actually do sentiment analysis as a goal as a researcher usually have. And as far as I know right now for sentiment analysis not only for digital media. So right now we can actually use those different techniques to actually transform or to input those more historical texts into the algorithm to identify those different machines to help it to learn from those historical context. For example like movie reviews or writer's expression of sentiments in novels. And I just read one article about like model the movement of narratives just within one story to see how the plots actually change across the whole book. So it's actually interesting to see right now sentiment analysis is not only used for those big data approaches but also it's actually being applied to more historical context. And one of the common methods we usually use to perform sentiment analysis is machine learning. So what is machine learning? And I think like one sentence to describe it is the extraction of knowledge from data. So basically you go through the data and you try to get out of the data. So you define it like more clearly is data machine learning is actually about the construction of algorithms that can actually learn from the data. And every time when you feed with different data with different learning processes it will improve with the experiences every time. And the goal of doing machine learning is usually like on the two different levels. The first one is to mix classification. So what I'm talking about makes the classification is when the machine can recognize and characterize the attributes from data sets. And those data sets can either be visual data or be measurement data. And the second goal usually that's associated with machine learning is to make the predictions. So the machine algorithm going through the data in order to find the patterns and then to make the predictions about the values. And one of a good example is if you want to know how much your house is worthy. You'll probably input all of historical data about the house sales in the past years like a GDP, how does the whole economic market changes with all those different data in the predictors being inputs. The algorithm is able to learn the patterns and actually to predict the value for you. So that's like a two different goals associated with machine learning. And there are several ways of doing machine learning. So you'll probably get a little bit more technical here about try to explain it in the lay people language with my best efforts. So the first type of machine learning you can see we usually use is supervised learning. So when you think about supervising it means there's a supervisor, so there's a teacher. So there's a teacher there to actually teach the algorithm to analyze the data. So which means when you give the data to the algorithm all those data are well labeled with correct answers and correct goals. So the algorithm actually knows what it wants to find out because it has a construction, it have like a textbook there to follow. So there's like a supervisor there. And the second type of machine learning is usually what we do is unsupervised learning. When you think about unsupervised it means there's no teachers, there's no instructions. So whatever you just feed the data, those are not labeled. You just feed the data to the algorithm. You hope the algorithm is going to learn by itself. Right, so think about if I give you one foreign letter you will be no way to understand it because you don't know the language. So it's the same for algorithm. However what the computer scientists expect to see is if you feed tons, like thousands, millions of letters even they are all in foreign language to the algorithm it's actually going to realize the pattern through those tons of data. However there is like, although there is like no instruction or like no textbook there to guide the algorithm it should be able to find out a pattern by itself. So it's what we call unsupervised learning. And the last one is reinforcement learning which is very similar to unsupervised learning. The only difference is every time the outcome is being graded. So if the machine is actually making a right guess, right move or right decision it's going to be in from the algorithm is actually doing the right thing. So it's actually going to learn from all those different decisions every time with the researchers actually telling the machine you are actually making a good move or making the right decision. So there are loads like a different type of machine learning processes. And here I want to show you one of the examples that you can actually apply machine learning to do the sentiment analysis task. Here I'm going to show you what we just talked about earlier is a supervised learning based method. So it's using the supervised machine learning methodology to actually to do the sentiment analysis. So there are actually lots of commercial tools that can actually do the same for you right now. And what it does it can think about it as a two processes, as a two steps. So for the first step is actually the job that's going to be done by human coders. So the humans, the human coders are actually going through a smaller set of documents like with the goal of the interest, a research interest. For example, I'm interested in an earlier case of a nuclear power. So I want to actually see how a smaller set of, for example, the tweets, actually the human coder are going to analyze the tweets and then to classify them into those specified sentiment categories like positive, negative, neutral. And what human coders do here is once human coders are actually training the data sets here, it's actually teaching the algorithm, right? So we talked about a supervised learning is you actually have a teacher, you provide the instruction for your machines. So here the human coders are actually the teachers here to provide the instruction for the algorithm to learn. So the supervised sentiment classifier, the algorithm is actually going to pick up what human coders are doing and then to go into the underlying patterns. And what it will do is actually it's going to apply what it's learned from the human coders to apply the learned patterns to analyze the rest of the big data. So because it's actually one of the, so think about it as a two process with the humans as instructors teaching the algorithm, teaching the machine to learn from us to actually analyze the data in a way we expected. And one of the good advantages of using those supervised machine learning based methods to do sentiment analysis is good for several reasons. One of the reasons because in a digital media world, we as human beings, we are not able to analyze millions of data, which is not possible. However, with the algorithm, you are able to do this more efficiently and more reliably. And one of the good things about not just relying on the machine to actually discover all the patterns on its own, but with the humans come in as an instructor to teach the machine, is the machine can never do the same thing as humans do. We have to admit it, it's like humans can actually easier to analyze the underlying patterns within, like linguistic patterns within those texts. So you are actually going to have a latent validity if you actually have a human coder to play a part in the whole analyzing processes. So supervised learning method is actually one of the emerging methodology at least in the field of communication. People are using this for sentiment analysis purposes. And I think some of them are using topic modeling as well for different purposes with the advantages that actually combine higher reliability, better efficiency, and also reserving the latent validity. So hopefully through this case to give you a better idea of our machine learning and also sentiment analysis. Thank you. So in trying to talk about metadata in a non-technical and very visual way, my thoughts immediately turn to Elvis. So I'm just warning you, there's gonna be a lot of Elvis ahead. Mojo Nixon had this awesome song, Elvis is Everywhere. I would say really metadata is everywhere. And metadata has been everywhere just in some of the topics that we've actually already talked about without actually bringing that forward as a term. So it's also something that you create on a daily basis. Who are you sending emails to? Do you have the location button turned on with your phone? You're creating a picture about you in terms of descriptive information with just how you're interacting with the digital world. I'm gonna go through different ways that metadata can look. It can look like this. This is a mock library card for this awesome movie called Bubba Hotep, which is where Elvis is in a retirement community fighting an evil Egyptian mummy. I encourage you to check that out if you haven't seen it yet. But in our library card catalogs, we actually still have a lot of data structured just like this. So for those of you who feel nostalgic for library cards, they haven't really gone away. They're just baked into many of the underlying systems that libraries still use today to some degree. And now we'll take a look at something else. So this is a comma separated values file of I think the top hits from Billboard from 1957. And so this is an example of metadata that you could bring into a software program like Excel and then manipulate it or augment it or switch it around. In other cases, metadata looks like this. We saw some XML earlier in that TEI example of markup for a poem. This is RDF XML and it is specifically dealing with the topic of Elvis sightings after death from the Library of Congress. So often when we're building digital humanities, exhibits representing things digitally online or in our digital collections in the library, XML is in the guts of the systems and knowing how to manipulate it and interact with it could be through an API is a great way for just leveraging that information. So some broad definitions. This is more from a librarian perspective and you can have multiple definitions of metadata but just that it is descriptive information. Some people say data about data. Sometimes it's useful to break it up into broad categories like knowing what the subject is of something, an author, a title, having a unique identifying number to describe, to associate with your digital object so it doesn't get lost or misplaced. And then thinking of things like file size and type and ways that you might be storing your digital metadata. And so here for the stamp, we have all of this descriptive metadata describing the image. We have the authors, we have the date issued. All of that can come into play if we're building a fabulous online digital Elvis exhibit. We talked about Spotify or Pandora. So again, this playlist will let me know how popular certain songs are in the descriptive metadata there. Metadata can be structural. So I'm leafing through this awesome FBI report about Elvis that's been redacted in certain places. But the thing that lets me page forward and backward and locate, you know, this is page one where he's talking about his corvette and then go to the next page that can be accomplished through structural metadata. It can also be administrative. So I have two different ideas that I want to refer to. I want to refer to the idea of Elvis's ghost and I want to refer to that as totally distinct from the actual person Elvis. And, you know, like many things, the Library of Congress has me covered. Those two concepts have unique identifying numbers. So that can also help me if I want to, you know, do some awesome digital stuff with Elvis. We've mentioned already mappings. We could do geospatial metadata and map our Elvis sightings on a map. So we know where to go if we want to find, you know, the spirit of the king. We can do topic modelings as well for selected Elvis lyrics and just find out how often he talks about being lonely, which is fairly often poor Elvis. So just as an aside, this is a sentiment from Jason Scott where he said metadata is a love note to the future and that's something that I've really taken to heart in the kind of work that I do. Without metadata, you can't find your stuff. You can't collect your stuff together and you can't tell people in the future what you're thinking about, you know, a research data set that you might be using. So if you're taking care of your metadata and you're publishing it, you're doing something really good for the scholarly community as well. So thanks very much. And if you ever want to talk about metadata, you know where to find me. Link data. All right, so I'm going to be talking about linked data, which is kind of taking the metadata that Anna was just talking about to the next level and being able to link it together to create a great big web of data. So if we just want to take a look at a simple webpage, we as humans can look at this and we see, okay, there's a great big image on there with a bunch of birds and some other animals. We can see there's a header with some information under that to get some context about what the digital library is. We can see on the right-hand side, we have some contact information, a phone number. There's links to other webpages. We can see where those links are going to take us. But then if a computer is looking at this, all it's really seeing is that there's a great big box, it's an image. There's text, text, text, text, text everywhere. There's all these links on the right-hand side. It has no idea where it's really going to go if somebody clicks on that. So in the early days of the web, Tim Berners-Lee, his major idea was to create a semantic web or a web of data where not only humans could consume the data, but also machines would be able to understand what it meant and could have some context behind it and be able to analyze that data. So to accomplish this, in 2006, he coined the term linked data. He said that there are these four principles that need to be followed in order to really make your data linked data so that computers can understand as well as humans. So the first thing is to use URIs or just unique identifiers for to name things. The things could be people, places, ideas, concepts, events, whatever type of data you have. So if you remember back on one of Anna's slides about Elvis as the spirit and Elvis as a living person, there were a couple of URIs on that slide from the Library of Congress. The second principle is to use HTTP. So basically just use the main protocol that's used on the web to retrieve data. When somebody looks at one of these URIs with the HTTP protocol, you need to provide useful information for them using certain standards. And then that useful information needs to link to other useful information so the person or the computer can get more context behind what this is really looking up. So the most common way to think about linked data is through triples. Basically a subject, predicate, and object. You can think of it as a simple sentence. This theme's author is Brown Stoker. So in that third line there, you can see if we break that sentence up into the subject, predicate, and object. And then the bottom example there is if we represented that with some potential URIs for that. Another way to think of linked data is through linked data graphs like this. So we have that same information, the subject, the predicate, and the object. But not only do we know that this thing has an author, we also have a lot of other metadata that we know about this. So we can start linking things like the publishing day or the title of that thing. Not only do we know stuff about this thing, we also know stuff about the author. So we can start linking other information about the author to their entity, whatever you wanna call them. So we have their birthday, we have other people that they may have known or worked with, we have other works that they may have authored as well as those have titles. So you can see as we add all of this metadata here, we're just creating one big web of data that's all linked together. If we wanna take that to the next level and have tons of data, you can see the more data you have, your web can get huge. This here is showing around 400 different data sets and how they're all linked together. All of these data sets have been made available as linked data. Some of them are relatively small with thousands of triples. Some of them are huge like DBpedia, which is the linked data markup of Wikipedia with trillions, if not even more triples that are all linked together and other sites are linking into that. So the theory behind all of linked data is great, but what does that really mean? So I have a few examples here of different sites that are using linked data that if you wanted to go and check them out, you could go and see what linked data looks like in action. So this first site is linked jazz, which is taking a lot of information about different jazz artists and linking that together to see relationships between different people to see who may have worked together, collaborated on different albums or played in different groups together. Another example is the Civil War on the Western border, which is linking people, events and places from the Missouri-Kansas conflict from the 1850s and 60s. So you can create visualizations and see how they're all interrelated there. We have timelines and maps on this site using their linked data. One of my favorite things about this site is one of the relationships that you can get at is a relationship between two people. This person shot this person. And then another site is Ochre, the online coins of the Roman Empire, which is just pulling a bunch of structured data together from several different databases and linking that all together to provide more context around these Roman coins. So this last slide here, several years ago, Anna and I created a website for a Utah Library Association conference presentation about linked data. If you wanted to learn more about linked data, this would be a good place to look at. There's information about different library projects on here, and there's also an exercise where if you wanted to get some structured data like linked data out of the Digital Public Library of America's API, you'd be able to go in there and follow the steps of that. So yeah, that's a basic overview of linked data. I hope you guys are enjoying this as much as I am. This has been really great. So let me go back to my PowerPoint. OK, good. That's showing up fine. So I'm going to talk just for five minutes and give a broad overview of the topic of data curation. So if you've been around libraries at all in the last five to 10 years, you've probably heard this term a lot. Even if you haven't been around libraries, you've probably heard the name data curation a lot. And what it basically is, it's just a fancy term for the care and feeding of your data or your data sets. The way I think of it, it's adding value to data, either by creating great metadata while you're collecting your data set. It's taking certain steps, doing best practices and activities to make sure that your data is not only understandable, but that it's actually accessible and available into the future. So data, of course, can take a lot of different shapes. If you're in the sciences, data might look like computational models. It might look like simulations. It might be laboratory notebooks. It could be biological specimens. If you're in the social sciences, your data might be survey data. It might be interview data. It might be ethnographies. It might be all of Twitter. It might be your data set. If you're in the humanities, it might be things like records of human history. It could be photographs. It could be newspaper. I mean, we've heard today about a lot of different types of data. So the way I like to think of data is it's evidence for your research. So whatever you're doing, that is your data set. And data can have enduring value, sometimes even more so than the publications that you produce as a result of the data. A good example of that would be the US Census. I mean, this is data we've been collecting since 1790. And because it's well-maintained, and well-described, and managed, we have an excellent view of what the demographics of our country have been from 1790 to today. Now, unfortunately, most data sets are not managed and maintained in that same way. And I think most of us do much smaller projects in the US Census, and we don't think about our data in the same way. So data curation, like I said, it's really this set of practices and activities you can do to make sure your data is valuable into the future. So what might those activities be? I just highlighted a few of them here for the sake of time. But here are things that people do to quote-unquote curate their data or to make sure it has that value. The first thing people often do is creating a data management plan. And that's something you do at the beginning of your project where you say, this is how I'm gonna manage my data, this is how we're gonna share it, how we're gonna name our files, achieve version control, and how we're gonna share it at the end of our project. Probably one of the most important things you can do is capturing sufficient documentation during your work so that if anyone else wants to look at your data, they can actually understand the steps that you took, your methods, that kind of thing. That way if they wanted to validate your research or verify it or even replicate it or build upon it, that documentation is in place to do that. And as we learned with metadata, even if they just wanna access your data, you have to have appropriate metadata to do that as well. Another data curation step is converting your data to open source formats or migrating it to new formats. So the data that you might have today, let's say you've done a survey and you have your data sitting in Excel. That's great, and Excel might be around in 20 years, but it's very unlikely the version of Excel that you're using is going to be the version that's around in 20 years. And it's also likely that we might be using something completely different to look at that kind of tabular data. So what you would do is you would export your data out of Excel into something like a CSV file, something that is open source and something that we can actually look at in 20 years and import back into whatever program we're using then. Another thing you can do is sharing your data in a repository. So we are busy people. A lot of us, we get very excited about our project and we keep it on our laptop or maybe we keep it in like a shared drive with our colleagues. But at the end of our project, we tend to kinda move on to other things. But before you do that, a data curation step would be putting your data from your project into a repository so an archive or a library could care for that data and add metadata to it and add things like digital object identifier so you can link your data to your publications, for example. And then lastly, another thing you can do is assigning rights to your data. So one of the problems with data management is that people don't know how they can use data. They might have access to a data set but not know if they can combine it with other data sets if they can use it for commercial purposes. So doing things like assigning clear rights to your data would be a data curation step to make sure it's valuable and usable into the future. So sometimes people don't associate data curation with the humanities. They think of really large projects, things like maybe the World Wide Telescope or the Human Genome Project. But in reality, as I hope you've kind of guessed, like you can use curation for any type of data. And so I'm gonna look quickly at this one example of humanities data curation that they've done at MYTH, which is the Maryland Institute for Technology and the Humanities. Let me pull this over. So over at MYTH, they've been doing projects, digital humanities projects for a very long time. They hold a lot of events. They do things like podcasts and webinars. One of the things they've done is they've collocated all this information, all the products of their research onto this website. So rather than letting individual researchers maintain all their own projects, you can actually go on the MYTH website. I can't see my, oh, anyway, I can't find my little cursor. But anyway, you could go on this website. You can see that they've captured the projects. They've added sufficient documentation so you could find them. You could search by things like the type of project, by the topic, by who sponsored the project, by the year that it was produced. This is data curation for the humanities. And I know that a lot of us are working on this digital matters lab, hopefully collocating the work that we do in one place would be a way that we could curate our data as well. Okay, so this, I think that in nutshell is data curation. So now I'm going to take off my presenting data curation hat and come back to my host of this event hat. And there's two things that I wanna talk about before we move on to any Q&A. The first, like David already said, is we do have a digital humanities symposium coming up. I really encourage all of you to register. It's $50 if you're not a student. It's $25 if you are a student. We are currently doing a call for proposals until December 1st. So if you have a project related to digital humanities, I very much encourage you to do that. The second thing I wanted to do is you guys have your cell phones, right? Cause we all can't live without our cell phones. Great, we're gonna do a workshop next month as well. And we're gonna do a tool overview. So you've heard about a lot of terms and topics related to digital humanities today. Next month we're gonna look at some of the tools because you might wonder how you achieve some of the things we've talked about today and what might be the appropriate tools to do that kind of work. So if you have your phones, what I would love you to do just because I'm thinking this might be our primary audience here, I want you to text, first of all, send a message to 37607. We're gonna do a quick poll right now. I'm gonna collect some data and curate it somewhere appropriate. So 37607 and then you're gonna text, it's my name except for the NGS. So Rebecca, C-U-M-M-I, 531. This is how you enter the poll. So I'm gonna wait cause I see you guys looking down. Do this first. And now we're gonna go to the poll. What I would like to know and we activate this, I would like to know what digital humanities tools that you all would like to learn about. Just to give you some ideas of things you could put in there. OMECA, Scaler, Zotero, Mallet, Neatline, Chronos Timeline. What other great tools are there out there? Anyway. Gephy, okay. Go ahead and start typing those into that thread so I can learn what people wanna learn about. Thank you so much for participating by the way. This is a really helpful test. Yes, you can put in as many as you'd like. And actually I'm gonna put in one too. Yeah, I mean, let's see. Even if there's something you don't know the tool but you know the thing you wanna learn, feel free to put that in this too. So if you're like, man, now I wanna know about topic modeling or I wanna know about sentiment analysis, feel free to put that in, that's fine. Open refine if you wanna learn how to clean your metadata, that's a good one. Just give this one more minute. Good to know. Okay, great. I also wanna just put it out there that if any of you know how to use these tools really well and would like to teach them, feel free to contact me because we'll be looking for people next month to teach them. Okay, great. The next poll we're gonna do, this is my first time using this too, so hopefully this works okay. Thanks, yeah. So the next question I have for all of you is what kinds of activities or work do you see happening in the Digital Matters Lab? And how would you plan on using a digital scholarship space? So this is brand new in case you guys didn't know. The Digital Matters Lab is a collaboration between four colleges. This is what we're calling the pop-up space and we're conceptualizing what we would like the space to be once we've got the full, robust digital scholarship lab. So if there's something you'd like to see us do, please text that information and I will do the same. This is kinda cool, isn't it? It's like data on the fly. Okay, if you keep putting stuff, it'll keep populating and so that is great. But I do wanna just come back and thank all of our presenters. I think you guys did an amazing job. One more round of applause actually just for all of you guys, that was great. Okay, and now I'm gonna open it up if anybody has any questions for any of these people, please feel free to ask them right now and we can just pull them back up as the questions or you can answer from your seat. I hear something you mentioned, Rebecca, which is they have different objects of study, which might produce different forms of data. But I'm wondering if there's a better distinction that might be drawn just to help me as I think through this way to see the boundary to the extent that it exists between the social sciences and digital humanities. Yeah, that's a great question. I think that the one major distinction that you could think about in terms of differences between digital humanities and the social sciences that there's, I suspect, in the social sciences more of an empirical bent in which the data and the tools are used to validate some of the hypotheses. Whereas in the digital humanities, I think we acknowledge that we are pretenders, we aren't social scientists, so we use those tools as a means of problematizing questions and leading to further questions rather than validating any hypotheses that we might come up with. Thanks a lot to everyone. Hello? Yeah. Thanks a lot to everyone for the presentations. I really learned an awful lot, which is good, because I have a data visualization project going on right now and so I feel a slight bit more like I know what I'm doing, maybe. But, and that's one of the things I would come to the lab for as well. But I have a question that may not be terribly well articulated as a question, but it has to do, it's related to something that you mentioned, Rebecca, which is you give the example of how folks might be, let's say, using Excel to analyze data and then saving the data in Excel-formatted spreadsheets and how in 20 years we're gonna have problems with versions of Excel or maybe new companies that are also producing proprietary software that has its own formats and things like that. And one way to think about that is that, well, it's obviously inconvenient and it poses a challenge for our own data archiving, but there are other implications as well, it seems to me, obviously, of relying on various commercial products for, whether we're relying on them as sources of data or whether we're relying on them to help us analyze data. Another example might be, let's say, Twitter or even Facebook. I mean, a lot of people are interested in mining data from social media, but ultimately, a lot of these social media platforms are effectively advertising platforms. They're just kind of using us and using our social media interactions to mine data for advertising, for example. So, and Twitter seems to be a company that's at risk. I mean, it might be bought, it might, there's no telling what happens. So I'm just, again, there's no really well-formed question here, but I'm just wondering if maybe a couple of you, as you're interested, as you're able, talk through some of the implications of that really complex collection of interests where we really want to think about open source methods of analysis. We want to think about open source platforms, but a lot of the data we might be interested in and a lot of the tools we might use are proprietary, are connected to commercial interests that may have their own interests that significantly diverge from ours, if that makes sense. Yeah. Yeah, sure. I think there are any of the other librarians that want to speak to it as well. So I don't have, obviously, the answers to these questions, but I think that you hit on several important points. The first one I heard that stood out to me is digital obsolescence, the fact that we keep these things in proprietary formats that have many dependencies. They have software operating system dependencies. You know, we are currently working on a project trying to save video game data because we have a department here on campus that's the EAE program, Entertainment Arts and Engineering, and their theses are video games. So basically, one thesis might have over 30 different file formats, and you need to preserve what would be considered the thesis. You have to preserve all of it. Well, libraries don't currently have the capability to do that, although we're working on it. Some of these problems I think, you know, we are starting to get a handle on. I think that we feel pretty good about our ability to save certain things, like PDFs, TIFs, but other things, it's still kind of an open question. I will say that people smarter than me are working on it, so that's good. Other questions like things like the Twitter archive. I mean, Library of Congress is preserving the Twitter archive right now, but they have problems with that. You know, how do you create metadata? Do you create it at a tweet level? Do you wrap 140 character thing in 500 words of metadata? No, you can't do that. And so sometimes, you know, when we try to solve these problems, we create even bigger problems. Did anyone here, let's see, so we've got people from metadata, metadata, IT services, IT services, anyone wanna speak to this? Sure, this is Harish Maranganti. Yeah, thanks, by the way, for all the wonderful presentations. Going back to your question, digital preservation, I mean, that's something that we are trying to deal with here in the library, and one of the challenging things is to think about what to preserve. Obviously, we cannot preserve everything that's being produced digitally right now. So do we let the process of natural selection help us dictate what we get to preserve in the sense that if we have, just to backtrack a little bit, I think the question right now is not a choice between open source or vendor systems so much. It's mostly about the community-driven approach, right? If you have a vendor system that's pretty common in majority of the universities or institutions, and if enough people use it seven years down the line, even if the vendor is not willing to support the product, is there a community around the tool that is willing to put their resources together and move that forward? So in the library world, we are talking more about community-driven approach rather than open source, where in previous years, you had one institution take the lead, develop something, release it as open source in the hope that others would adopt it, but now the approach is get a community together, talk about the requirements, and then see whether we could develop something which is useful for everyone, which speaks about the sustainability of the approach and quite a few things too. So I think that that is where we are digital, again, digital preservation is a challenge. What we try to do whenever we talk about products is to think about an exit strategy. It's not enough to use the tools. Can we get the data out? Can we get the data out in a way where, again, it would be useful for the future scholars? We have pretty good luck with the text-based system so far, but audio, video, it's a challenge because again, in terms of maturity of some of the standards, we are in its infancy, so there are not many standards out there that would do it, but again, we are involved in some national initiatives. We are part of what is called a digital preservation network. So we are at University of Utah as a charter member, so we are collaborating with 60 other universities to think about how do we preserve cultural material that might be useful 50 years down the line. We don't have all the answers, but we know who to partner with and ask the right questions at this point. All right, did anyone else have any questions right now? Okay, because I have a question, so I'm gonna pose this to some of our faculty members over here. I think sometimes I have a good idea of how the library fits into the digital humanity space. We do a lot of metadata creation, of course, preservation. How do you guys see libraries partnering with digital humanists to get this work done? Or do you see us as partners? Well, thanks, Rebecca. As part of my answer is specific to this institution because we are very interested in building a digital humanities program here and we are partnering with the library not only in terms of procuring the space, but in terms of pushing forward the field of digital humanities. So the example that I always give in terms of placing my value in the contributions of libraries is that I read this great essay by a librarian who took a lot of digital humanities scholars to task because we had been using the word archive kind of in a very cavalier way, not really fully understanding what that means in the long history of the ontology of archives in library studies. And so in some ways I feel like librarians are here to kind of keep us honest because we kind of throw these terms around, not really knowing what we're talking about and libraries can kind of keep us in check. But also I am quite interested in the problem of preservation and curation because as Rebecca's presentation pointed out, a lot of what we do is about not looking at, as sort of literary historians we look at the past, but we also have to look towards the future and we need to have very meaningful partnerships with librarians and their expertise and making sure that we can write those love letters to the future because otherwise all the research that we are doing can just rot in our digital sellers really. So that's one of the major components about sort of the cultural shift that we are trying to interested in pushing forward in digital humanities. And thank you, that was a great answer. I wanna add to that because Rebecca was asking what is the role of the library? And I'm wondering where do the students fit in, right? Because Hema, who is sitting over there and Abby wanna teach a class where we use digital humanities but we also wanna teach it, right? But I don't foresee myself teaching my students to do the text mining, to use these software programs that we both spent a week learning, right? In full on four hours a day. And I'm wondering where is that piece that will help us implement the digital humanities in the classroom? And I'm not sure myself whether it is something we can foist on the library, right? Because you guys are already busy with what you're doing. But it seems to me that teaching and learning or some of the pedagogy centers on campus should be involved in the initiative to bring that third piece in. So it's not really a question, more a suggestion maybe. So cheesy with this microphone. Excuse me. So I'm teaching a digital humanities class next semester and another lower division class called Literature by the Numbers. And in both of those classes, I want my students to have that kind of hands-on learning. So one approach that I'm going to try next semester is to have one day a week in the traditional classroom and the next meetup session in the Mac lab here in the library. And so I have things that I'm comfortable teaching, but I'm also hoping to kind of identify a network of expertise around me. And I'm new and maybe people already are aware of that network, but I think it would be useful for all of us who have that interest to say, hey, Lisa, would you wanna come to my class? Or hey, Dave, would you wanna come to my class and kind of give a spiel about something? And I think that would be very helpful to just have a kind of larger network where we knew that we could kind of call on each other and build off our expertise. And so that's something that I hope that we can create. I think one of the goals of the Digital Matters Lab is the fact that many of us work in our colleges and departments, but this is something that can be interdisciplinary. It can be collaborative. And the fact that it occurs in the library, we do serve, I think, as a neutral space as well, where lots of people can come and use the resources and not feel like they're infringing on a different department. So yeah, I think that that's something maybe that we should make a focus as well to learn how to incorporate this into the classroom. So I see that we're just two minutes shy of 3.30. I thank you all for staying this late on a Friday afternoon. And unless we have one more burning question, we can probably make this the end. Anyone else, a quick question? All right, thanks, everyone.