Creative Capital: Collaborative Approaches to Developing and Supporting Digital Contexts for Hyperlocal Histories
Brown University, United States of America
This paper describes ongoing cultural initiatives in Providence, Rhode Island that focus on “hyperlocal histories” of these cities and communities: narratives, archival collections, and exhibits that take a close look at particular neighborhoods, buildings, parks, residents, political operatives, murals, artists, and activists (among other topics) over time. It will call particular attention to the ways these initiatives consider digital contexts for their work and what ideas of value seem to inform decisions to utilize particular tools, platforms, social media networks, forms of digitization, curation, and preservation. Providence is nicknamed “The Creative Capital” because it is home to an active community of artists and educators, many of whom focus their efforts on stories of the city. In 2019, a collaborative initiative titled “Year of The City” will document the varied ways local residents and cultural institutions think about Providence and the long histories of its twenty-five neighborhoods. I am particularly interested in when and how these images and perspectives on the city circulate online, materialize in archival records, reach local publics as well as national or global audiences. In looking at the uses of particular technologies, methodologies, publication platforms, and avenues of outreach and dissemination, I will highlight the obstacles and challenges facing creators of hyperlocal histories within and beyond the contexts of higher ed, and I will document forms of collaboration, pedagogy, and institutional support that can aid efforts in these local contexts.
This paper will highlight particular digital initiatives invested in the hyperlocal: developing work tied to “Year of The City,” the popular podcast Crimetown, the mobile phone tours offered by Rhode Tour, images and ideas of Providence on Wikipedia. It will also consider popular efforts that prioritize non-digital contexts: the Rhode Island Collection of materials at the Providence Public Library, the Dirt Palace feminist art space, physical tours of what was once the city’s Chinatown neighborhood, artistic interventions at Mashapaug Pond by the Urban Pond Procession. By bringing digital contexts related to technologies of tours, exhibits, archives, audio storytelling into conversation with non-digital projects invested in these same metaphors and methodologies, I will assess possibilities of remediation, alternate forms of dissemination and strategic re-use of digitized and born-digital materials, and possibilities for collaboration. And I will discuss the work of ethical and inclusive forms of collaboration, resource-sharing, and compensation that I have been part of in recent efforts to connect digital humanities practitioners at Brown University with community partners and practitioners. While Providence has its own particular challenges and contours, my hope is that a candid assessment of ongoing work related to hyperlocal histories will help conference attendees consider how their own research and resources might lead to generative partnerships with local practitioners and audiences.
Urban Panorama: t-SNE Street Feature Mapping Tool
North Carolina State University, United States of America
In the last two decades, historians have increasingly employed GIS to understand urban and spatial change. However, GIS approaches landscapes from above, reproducing the point of view of the planner. The Urban Panorama Project aims to introduce a new dimension in the historical assessment of how cities change. By shifting the focus from geometric parcels, as seen from the air, to images of streetscapes, as seen at the street level, we intend to move closer to the perspective of the people experiencing change in the space of the city as they traverse its streets and avenues. The Urban Panorama Project, therefore, is testing different computer vision and machine learning techniques to assess historical and present-day images of streetscapes to investigate urban change. One of these techniques, and the object of this presentation, is exemplified in our t-SNE Street Feature Mapping Tool. The tool allows the visualization of clusters of tens of thousands of building facades, or other desired street-level visual features, scraped autonomously from geolocated street-level photographic corpora. The system leverages two popular machine-learning techniques and open source software libraries. The first, YOLO, is a neural-network-based computer vision object detection system. We are able to train desired image classifiers (e.g. what does a building facade, commercial sign, or mailbox, etc. look like), by providing hundreds of example annotated images. Such classifiers are then used to autonomously mine photographic archives and extract these features (and their associated geo-location/temporal metadata), depositing the results in a database. The second, t-SNE (t-distributed stochastic neighbor embedding), is a machine learning technique for dimensionality reduction, useful for visualizing high-dimensional datasets. Unlike YOLO, it does not require human training—in our case we used the t-SNE technique to analyze all images in a particular feature corpus—i.e., all building facades, etc.—and autonomously cluster visually similar images together in a spatial plot. In this presentation, we will present the tool and its application to a corpus of geolocated historical (1920-1980) and present-day streetscape images of Raleigh, NC. The tool uses an intuitive thumbnail grid interface where features are selectable and visualized as a heatmap layer in an adjacent city map. The t-SNE Street Feature Mapping Tool will help researchers within the Urban Panorama Project to use historical streetscape images to understand the spatial distribution of urban features across the space of the city. With the development of this tool, we can tap into corpora of street-level historical images to understand the spatial distribution of streetscape features in a city and compare different time periods. Phenomena such gentrification, urban decay, the spread of architectural styles, the use of different materials, textual analysis of urban signs, social uses of public space, urban flora, etc. could be mappable through this technique.
Slave Streets, Free Streets: Mapping the Dispossessed and Un-Addressed in Early Baltimore
University of Maryland, Baltimore County, United States of America
Advances in historical GIS have made it possible to map the past in ways that would have seemed impossible a few years ago, georeferencing disparate maps in order to build deeply accurate visualizations and recreations. But what happens when we reach the limits of our information and sources? How do we map people who don’t have addresses?
This paper grows out of an effort to map the lives and locations of free blacks and enslaved workers in an immersive map of Baltimore, circa 1850. One of us is an historian of the 19thcentury Unied States, and the other is a photographer and animator. This deeply researched, detailed site Visualizing Early Baltimore, http://earlybaltimore.org is like a Google Map for the 19thcentury abd serves as a basemap. We want users and researchers to be able to walk down the streets of this virtual city and learn about the lives of people usually left out of historical narratives. Too, we believe that these maps will show the degree to which free blacks and enslaved workers lived in an integrated, rather than segregated world, a marked contrast to the deeply segregated Baltimore of today.
This paper will discuss our efforts to solve this problem in ways that have implications for historical reconstructions of other cities in the years before the mid-nineteenth century standardization of urban addresses. We have chosen to foreground our uncertainty, to show the dearth of information about black lives and spaces, even as we attempt to bring forward the names of individual enslaved people and free blacks.. Their places on the historical landscape can be visualized and contextualized.
Mapping City-Scale Reading Events: Geography and Sentiment of "One Book One Chicago"
DePaul University, United States of America
In The Bestseller Code (2016), Jodie Archer and Matthew Jockers argue that “while it does matter whether an author chooses a city or the wilderness, the specific city does not matter all that much when it comes to bestselling.” In this paper, researchers from the “Reading Chicago Reading” (RCR) DH project team will demo tools and methods for determining whether the settings of books do in fact have a measurable influence on reader popularity.
The RCR project studies the Chicago Public Library’s (CPL) ongoing city-wide “One Book One Chicago” (OBOC) program to capture correlations between circulation data and outreach programming, and to create tools and predictive models that would help librarians increase patron engagement. We will provide a walk-through of analysis and visualizations of CPL checkout data, associated social media, and community programming events since 2011. If literary representation of place and real geography have detectable links to one another, our RCR project data can capture the effect.
Using city-wide library branch data we have received from CPL, we first ask whether Chicago-themed OBOC selections check out at the same rate as non-Chicago-located OBOC selections. Initial results indicate a statistically significant difference in checkout numbers per branch even though CPL maintained roughly similar marketing efforts during each season.
Next, we will explain how real and imagined geography in texts do and do not directly relate, and how sentiment can link to place. Using Stanford NLP tools, we extracted Chicago locations in several recent OBOC works for which we have real branch-level circulation statistics and sentiment scores that have been assigned using automated methods. Our visualization reveals pockets of city space that consistently or exclusively receive negative sentiment scores in the selected books, and show checkout effects.
Finally, we will present spatial correlations between location-based sentiments and socio-economic, demographic, and crime statistics using American Community Survey (2012-2016) and CPD data. Initial analysis shows that most sentiments metrics, especially negative ones, are higher in areas where there is higher inequality (as measured by a Gini index 45%+).
Our analysis suggests that in some cases, and despite Archer and Jockers’ provocative claim, the particular city does matter when readers are in that same city and can recognize place names.