Computational methods include using techniques such as machine learning, graph inferences, and web development to reconstruct and communicate the social networks of early modern Britain from about 1500 to 1700.
Carnegie Mellon University and Georgetown University researchers, including Christopher Warren, Daniel Shore, Jessica Otis, Scott Weingart, Cosma Shalizi, and Raja Sooriamurthi created this digital humanities project to look at big historical data to see how often names are mentioned together in the history of scholarship as a way of modelling social networks.
Their work is published in the July 2016 edition ofDigital Humanities Quarterly.
"Our website allows scholars, students, and citizen humanists to improve the network - that is, add relationships to validate some of the inferences that we've made, and in many cases to reject some of the statistical inferences. This means that over time we get a more accurate representation of the social networks of the period", Christopher Warren stated.
In essence, Christopher Warren and his colleagues take the history of scholarship in so far as it's been digitized and run it through algorithms to see how often any two names have been mentioned together. The machine learning aspect then finds ways to model these past relationships. The hope is to find a model that accords with what they've learned through years of study and helps extend their knowledge to new networks.
Trying to understand the historical context of the major literary and artistic works and ideas that emerged in the 16th and 17th centuries is no easy feat. The 200-year period that brought us the Reformation and the scientific method also brought us Hamlet, calculus, and the microscope.
"The only way you can understand any of these things is by understanding the context from which they emerge. If we want to understand how we got something like Paradise Lost or the separation of church and state, it's going to require us to pay attention to who knew whom and how ideas spread, and the ways in which our modern world is in crucial ways a function of historical social networks", Christopher Warren stated.
Take, for example, the relationship between authors William Shakespeare and Christopher Marlowe. People have long supposed that William Shakespeare and Christopher Marlowe existed in the same milieu and more than likely knew one another. But what scholars are finding now based on internal analysis of their work is that they were more than likely co-authors. In the new Oxford edition of William Shakespeare's complete works, which will be available this month, Christopher Marlowe will be credited as such.
This is precisely the kind of finding that can be integrated into "Six Degrees of Francis Bacon". As scholars find more examples of relationships, they can go to the website, add them, and see them integrated into the most current picture of scholarly knowledge.
"In our case, we inferred because William Shakespeare and Christopher Marlowe's names often appear near one another that they probably knew one another at a 75 percent probability", Christopher Warren stated. "Recent evidence seems to confirm this, so we can bump that confidence up to 100 percent and have an even better picture of the past."
A project like this generates tons of data.
So, in July 2016, Christopher Warren and his colleagues became users of the Extreme Science and Engineering Discovery Environment (XSEDE) to help them analyze the data and to expand their data sources.
"We've primarily been working with the Oxford Dictionary of National Biography (ODNB), which is the gold standard of British lives from the Roman Empire to the present. Much of our initial work is with that corpus", Christopher Warren stated.
But they needed more sources to verify the validity of the relationships they had found. They are now expanding with help from XSEDE's Extended Collaborative Support Services (ECSS) at the Texas Advanced Computing Center (TACC). David Walling, the ECSS expert at TACC, is helping them see if the process they used on the ODNB can be extended to other corpora such as historical journals.
"If we look at a large corpus of journal articles and we ask how often names appear near one another do we get a similar result as the ODNB or do we get something different?" Christopher Warren asked.
With the help of ECSS, they now have 15,000 people in the data base and on the order of 100 million possible relationships.
"Once you employ computational techniques you can start to assemble relationships at a much greater scale", Christopher Warren stated. "This is something no human could ever have in their head. By putting this together and making it available for the scholarly community we hope that we're facilitating a new way of doing scholarship that allows for a full appreciation of these historical networks."
"We couldn't develop the project in the direction that is most useful without ECSS", he stated. "ECSS allowed us to extend our early work and move forward with it rather than spin our wheels. I can't say enough about the impact that the ECSS programme has had for the project."
Although most advanced computing is used for the hard sciences like physics and chemistry, this project is a unique collaboration between computer scientists and humanists. The project couldn't progress without the help of David Walling, who is adapting code in R and deploying the website code onto virtual machines. The domain expertise of both computer scientists and humanists was central to the success of the project.
"Right now, part of my role is to take all of the R code which produced the network graphs that are visualized on the website, and make that R code usable with new datasets", David Walling stated.
R is a programming languagewidely used among statisticians and data miners for developing statistical software and data analysis. In addition, David Walling is working in collaboration with XSEDE Campus Champions Fellow Xinlian Liu of Hood College. Their current focus is on applying the workflow established by Christopher Warren's research to a set of 450k+ articles from the JSTOR digital data collection.
This was Christopher Warren and his collaborators' first foray into XSEDE and ECSS. He has an interesting viewpoint on it.
"I'm not sure we would have become involved in XSEDE if it were not for ECSS", he stated. "The collaborative support model was attractive because someone with my background and training was intimidated by the prospect of using supercomputers. Knowing that there was a process to get our team up to speed was incredibly influential in bringing us on board."
David Walling is fairly new to this type of deep collaboration as well. "It's relatively early in the project so I'm still learning the specifics of the algorithms, how they work, and why they are particular to this data set", he stated. "To me, the interesting parts of the project are the machine learning algorithms and the statistical analysis that goes into building these social networks from text documents. I'm excited to have the chance to dig deeper into the consulting roles of different groups getting to see what people are actually doing with our systems."
Since the website went live in September 2015, there have been more than 50,000 hits and about 500 active users who have created accounts and are contributing to the picture of the past. In addition, this research project is being taught in classrooms across the United States and it's been the focus of several workshops.
"It's been a successful launch and one that we hope the ECSS and XSEDE programme can continue to help support", Christopher Warren stated. With a grant from the National Endowment for the Humanities, Christopher Warren and his colleagues are planning to re-design the website and find a long-term home for the project.
"One challenge that visual projects like this face is preservation", he stated. "Libraries are really good at preserving hard-copy books but digital artifacts like websites can fall obsolete very quickly."
The ultimate goal in redesigning the website is releasing the code to the scholarly community so other people can build and create similar networks for other time periods.
"We're doing a lot of documenting of the existing code to make it more user friendly, helping anyone who might be interested in doing something similar", Christopher Warren concluded.