Welcome to the New Books Network. Hello everyone, welcome back to Noob's Network. This is your host Shu Wan. Today I feel very happy to invite Dr. Sanders to join us to introduce her fantastic book. It's called Visualizing Historic Fragments. So for first thing today, I want to invite Dr. Sanders to introduce herself to us. Hi, everyone. I'm Dr. Ashley Sanders. Most recently, I was the vice chair of digital humanities at UCLA. I am now a data scientist working with small businesses.
using all the same skills that I present in my book, Visualizing History's Fragments, both to continue my own historical research and to help companies. Okay. Thanks so much for your answer. So for next question, because I'm, you mentioned like your topic, the topic of book is about visualization for history and of history. So I'm wondering because.
I think it's very, it's not so common, like visualization today, even today, it's not so common tool for history, but why are you taking interest in that? Yeah, so it sort of developed organically. I started with a question based on a document I found in the archives, and this is usually where we start, right? So I found a memoir from the last Ottoman governor in Algeria.
who was mixed ethnicity. He was both Ottoman, Turkish, and Algerian. And I began to wonder how representative he was of other governors. And of course, questions of representativeness lead us into statistics. And with that, we can also explore our data once I put the data set together with visualization. So it sort of just developed as I dove into that. First, then let's tend to your book.
So for the first question, I'm wondering about the history context of ultimate algebra, as well as, I mean, your fashion, as well as the nature of your book, your research here. the impact of the coronavirus on the availability of source material and how silence entered our archive.
Yeah, so this is a great question. I think it touches on commonalities with a lot of other colonial empires and the study of colonies, especially in places like Algeria, where it was sort of a palimpsest of cultures and of colonial empires. So one of the challenges I faced with the study is that after the French conquest of Algeria, many of the documents were lost or stolen or destroyed. And so that led to a lot of silences and gaps in the archive.
which is why I began to turn towards computational methods to interrogate the data I did have and to also uncover what the silences in the archive actually mean. Thanks so much. Then let's turn to the next question. That's chapter 3. And for this chapter, I'm wondering how to, because I think personally speaking, I really like your idea. But anyway, so I'm wondering how to, how do you employ topic modeling to summarize the primary source?
It can ban the task analysis technique with sentimental analysis, another very popular technique for, I mean, for data analysis to integrate the thought to any underlying bias. Yes. So I'll explain a little bit about what topic modeling is because it might not be commonly used. So it's a method from natural language processing, and it works best when you have a large corpus of documents. But I did this with a smaller set of documents that worked quite well.
And what it does is it automatically detects themes and it... groups, words that appear, and statistically significant patterns together to form clusters. And then we, as human researchers, take a look at those clusters to identify the themes and topics that they seem to describe. And then what I did is I used a hierarchical approach, which means that I developed models of different sizes. So a model of four topics and a model of seven topics and a model eventually of 20 topics.
What this did is it allowed me to see in greater detail how these themes emerged. at more granular levels. So the topic or the model with four topics gave me kind of just a general sense of what this collection of documents was really about.
And then as I built larger models, that allowed me to get much more detailed. And I could look at, for example, the roles of... people who had been marginalized in the society, the roles of women, the roles of Jewish people, the roles of other kind of ethnic minorities in Algeria. Then the next step after developing that kind of hierarchical model to get a detailed sun.
of what was happening in these source materials I wanted to know if there was bias in the materials that I hadn't caught when I read through them and so I employed sentiment analysis but not on the original documents. I used it with the topic model that I built. And what I uncovered is, you know, I thought for Algeria and many of these writers were French, I thought, well, maybe there might be an anti-Arab or anti-Turkish sentiment.
And what I found was actually an anti-Semitic sentiment that I had completely missed because it only appeared in fragments of the materials, just a few phrases, a few stories here and there. And so the next question I asked was, well, is this a bias of the authors themselves or was it a bias operating in the time period that they're recounting? So in the Ottoman period in Algeria's history.
What I found was that it was a combination of the two, but also predominantly some really unfortunate events that happened to Jewish people in the history that the authors were recounting. So using this method, I was able to uncover things that I hadn't caught, even though I had closely read all of my sources. Thank you so much for your answer. So now let's turn to chapter 4.
For this chapter, I'm wondering about how archival science may speak in the detail of the considerations and assumptions underlying the dataset construction. Yes. So as I constructed the data for this study, I needed to think very carefully about how I was classifying individuals, what categories I was using, how I was describing them. And I'll provide one example. One of the most difficult categories to construct was that. a variable I called ethnicity, which I have since changed to origins.
In my original version, in the version that you would read in this book, I call it ethnicity because I didn't know what else to call it. And with this variable, I was trying to get at the sense of where did these people come from? Who were their families? Were they born in Algeria? So were they kind of native to that region? or were they coming from elsewhere? And for those who were coming from elsewhere who were part of the Ottoman administrative elite,
I didn't break that down further with this variable. I used other variables to classify what cities they were coming from. Were they ethnically Turkish? Were they something else? Because in the Ottoman Empire, many of the people who served in the administration were not native Turkish people by ethnicity. They were often Christians from other European states that became part of the Ottoman Empire.
So I lumped all of the Ottomans together in this variable and I used other variables to dig down into that detail and that richness of their backgrounds. But that was one I really struggled with. And I explained this in the text. how I struggled with this variable and defining the different classes and why I structured it the way I did.
And with my documentation with this data, I also provide that description so that anyone using this data will know how and why I created the classes I did so that as they're doing their own analysis. they can keep that kind of cultural sensitivity at the forefront and understand how this variable operates, how they can describe it, and how they can use it. Yeah, thanks so much for the answer. So for the next question, I'm wondering about data aggregation and exploratory visualization.
Yes. So when we put a data set together, sometimes... Our questions lead us in slightly different directions than the questions that motivated the data set construction initially. And this is especially true if we're working with data that someone else created. And so if that's the case, we often have to use aggregation techniques that will help us answer these other kinds of questions. I'll provide an example.
So I was curious about how long governors were in office. In general, in the Ottoman Empire, they were in a governorship for one to three years before they were moved to a different province. But what we found in Algeria is that governors were in office much, much longer. And so one aggregation strategy I used was to group.
the amount of time that governors were in office and create what's called a histogram to visualize how many governors were in office for, say, zero to one year, from one to three years, from three to five years, and so on and so forth. And that gave me a sense of kind of the distribution of experiences of the governors in Ottoman Algeria. And then I could compare that with the experiences of governors in other locations.
So this is one other thing I'd really like to point out about visualization and about computational analysis. we can stay focused on our individual study, but it also becomes really powerful when we can compare what we're seeing in one region or with one group of people to the experiences of others. That helps us create a broader context. to understand what we're seeing in our visualizations and statistical analysis.
Thanks so much. So now let's turn to the next chapter. So for the next, I'm sorry. For chapter six, I'm wondering about how to employ descriptive statistics to explore how image characterized. representative governor's channel. Yeah, so this was an interesting chapter. This chapter talks about how we use descriptive statistics.
to understand the question of representativeness. This was the question that animated the study to begin with. So it was really important for me to spend time, dive in and explain how I... how I define representativeness, both in history and in statistics, where the crossover is. And then we dive into descriptive statistics to get at this question. So I looked at things that you might remember from years ago, mean, median, and mode. to understand, okay, what was sort of the average?
governors experience like? And did this change an historian? So the question of time is really important. Did this change over time? So I broke my data into two different kind of sub data sets. one for an earlier time period, and then I noticed through my explorations with visual analysis of the data.
that there really seemed to be a pivotal moment in 1792 with the death of a long-serving governor. He was assassinated. And after that, governors tended to be in office, it seemed like, for much less time. And so I wanted to explore, is that really kind of a turning point in the history of this region and in the governor's experiences? And by using descriptive statistics, I could show that governors who served before the state.
did indeed serve, on average, much longer tenures in office than those who came after 1792. And so there was a marked difference between these two time periods. And that led into questions of why, you know, what about this time period really caused this change? That gets me more into the historical side of things. But you can see how I'm using the computational analysis to drive me back into historical questions and questions that are significant to the social science analysis.
of this time period and the experiences of real people. Sorry, answer. I very appreciate that. And then for the following question, I'm wondering about, it's about chapter seven. I'm wondering about how to show a status test for independence, cooperation, and significance. maybe how this kind of task may be used to explore potential potential relationship in the data to underscore role of actor.
who often have our official Ottoman archive, especially local women and their family and Algerian political elite. Yeah, so often when we're doing analysis for humanistic or social science questions, we have... data that is categorical. And so this chapter really dives into how do we work with that categorical data to uncover patterns and trends.
For example, my question was about the origins of these governors and did that have an impact on their fate at the end of their tenure as governors? Because what I noticed is that... A number of governors were able to hold office for long periods of time and have a peaceful exit. Either they died in office or they moved on.
They went on the Hajj. There were many other outcomes for governors, but a number of them were also assassinated. It was a really violent time period. And so I was curious if their origins had any impact on their fate. And so I constructed a table that summarized. the different outcomes for different origins of these governors. And then I could do a statistical test called a chi-square test to see if those variables, fate and their origins, were correlated.
or independent. If they're independent, that would mean that one variable, their origins, had no bearing on their fate and vice versa. And what I found is that there was a slight tendency toward, especially for governors of Ottoman background or Algerian background, there were distinctions and differences in the outcome at the end of their tenure. What I found to be even more significant, though, in a statistical sense, was the difference in time periods that I mentioned in my previous answer.
So I was able to identify both through visualization and through statistical analysis that, yes, the difference in the time periods that I had noticed from early to late. was actually statistically significant as well as historically meaningful. Dear old work platform, it's not you, it's us. Actually, it is you. Endless onboarding?
Constant IT bottlenecks? We've had enough. We need a platform that just gets us. And to be honest, we've met someone new. They're called Monday.com. And it was love at first onboarding. They're beautiful dashboards. They're customizable workflows. But we're moving on. Monday.com, the first work platform you'll love to use. You just realized your business needed to hire someone yesterday. How can you find amazing candidates fast?
Easy. Just use Indeed. Stop struggling to get your job post seen on other job sites. With Indeed Sponsored Jobs, your post jumps to the top of the page for your relevant candidates. so you can reach the people you want faster. According to Indeed data, sponsored jobs posted directly on Indeed have 45% more applications than non-sponsored jobs. Speed up your hiring right now with Indeed. And listeners of this show will get a wonderful... Thank you.
support our show by saying you heard about Indeed on this podcast. Terms and conditions apply. Hiring Indeed is all you need. Thank you so much for that. Now, let's turn to the last question today. For this chapter, I'm chapter 8. I'm honored about the significance of aliens, the formation, slow marriage, and the role of local women in Ottoman Algerian society.
Yeah, so for this chapter, I used a method from network analysis. So I created a separate data set to uncover the social and political relationships between people from my source materials. So I was looking at the role of, specifically the role of women in Ottoman Algerian society. And what I discovered is that they were really the pivotal key actors. in this sociopolitical network.
And this was because Algeria was sort of on the frontier of the Ottoman Empire. Administrators who ended up there stayed there. They really didn't have a chance or the opportunity to move on to other more elite roles. What that meant also was that many decided to establish families there. And what I noticed is that those who married local women and into local families tended to fare better in office. They had better outcomes.
So the women's agreement to the marriage, which was crucial, most of these fathers would not have pushed their daughters into these marriages for political reasons. They would ask for their daughter's consent. to the marriage. So the woman's consent to a marriage was very important for an Ottoman administrator's health and well-being and their efficacy in Algeria because these women brought a wealth of knowledge and experience. They were literate.
They were numerate, meaning they could manage their household finances. They handled charity donations. And they were also often multilingual. So whether a woman was from a Berber family and spoke a Berber language or from an Arab family and spoke Arabic. They all knew to read, write, and speak Arabic, which was very important for local relations and even more important to speak.
Berber. And so they often served as sort of translators for their Ottoman governors and helped them with their administrations. I even ran across one story of a woman who was so revered for her knowledge of the culture and and politics in the region, that her husband, who is the governor, his counselors would go to her when they had questions about how to navigate this sociopolitical terrain. And she would serve as an advisor to the advisors for the governor.
So these women were super important in this region and in the society and indeed in politics as well. And all of this came out of my network analysis. I'm using math to uncover these key people within this network. So you can read the chapter for more details, but one of the measures I used was betweenness centrality, which helps to identify the bridges between different parts.
of a social network. And the women ended up being some of the most important bridges between different elements of this socio-political network. Thanks so much for your answer again. So now, at the end, I think I appreciate that you have already answered all my questions prepared for our podcast today. So last thing I want to do, just like I want to recommend my readers.
I'm sorry, my listeners should consider reading or buying a copy of Dr. Sanders' fantasy book, visualizing his story's fragments. So thanks so much for listening to our podcast today. Have a good day.