We are developing a methodology for exploring and finding meaning in large corpuses that contain images, such as archives of periodic publications. We focus this work on
Time magazine, and in particular on images of faces in
Time. We use computer vision analysis, combined with contextual research and methods from the humanities, to elucidate trends and patterns in the visual culture reflected by the publication. In particular, we are examining how representations of the human face have changed over time, and seeking relationships between the visual features we discover and their corresponding socio-political contexts. Specifically, we are interested in gaining insight about how the form and context of representations of women and ethnic minorities have changed over time. Our preliminary research focuses on the correlation between changes in facial representations in
Time magazine and the Women’s Liberation movement in the United States in the 1970s and 1980s. The main outcome of this project will be a meaningful and accessible web-based platform through which both researchers and the general public can explore the archives of
Time magazine to discover insights into our cultural history. We expect that we will be able to apply our methodology to any periodical publication, but we chose
Time because it stands as a record of the many pulses of U.S. and world politics and their intersections with American culture. We believe that because it is such a culturally important and ubiquitous publication much can be learned from these archives about how Americans perceived politics and culture throughout the twentieth and early twenty-first centuries.

Our methodology combines computational processes, such as computer vision analysis, with contextual research, such as the history of the magazine’s production process, as well as the cultural and political climate in which each issue appears. A brief summary of our methodology is as follows. Using the entire
Time magazine corpus (about 4800 issues spanning over 93 years), we are identifying and extracting every facial image within the corpus, and running computational analyses on the images to quantify their visual features (such as RGB pixel values). We are building a database of the images that includes their associated metadata (year, issue, page number), as well as the extracted visual feature data. Within this database, we are also including more detailed metadata for each image: the face’s gender, race, the context in which the face appears (ad or feature story), whether or not the face is smiling, and whether it is an individual portrait or belongs to an image that contains more than one face. In parallel to building this database, we are developing timelines of significant contextual information, which includes a timeline of the evolution of printing technologies used by
Time magazine, a timeline of culturally impactful geo-political events, a timeline of civil rights movements, and a timeline of women’s movements. Our image database will be connected to our contextual timelines with visual analytics. The visualizations we create will be interactive, web-based, and open to the general public.

We present here compiled preliminary results using our methodology and samples from our private collections of
Time magazine, along with a contextual timeline of the women’s rights movement in the US. In the work presented here, we used human labor to extract face images from sample issues and to tag each face image with the metadata described above. We are using the data harvested through human labor to improve our facial recognition algorithms and to train new algorithms to identify gender, race, smiling, and context. There has been a great deal of interest in sentiment analysis and facial recognition across academia and the general public, and we feel this project will allow us to examine how these complex categories interact with each other. For example, how do our understandings of race and gender impact how humans classify sentiment? How do these understandings impact algorithmic classifiers? This complexity is one of the primary motivations for developing a methodology that consciously moves back and forth between human and computer analysis.

The metadata extracted by human labor has been particularly insightful, especially when put into the context of our historical timelines. Specifically, we noticed an increase in the number of female faces in the 1970s, coincident with the many milestones in the Women’s Rights movement. Interestingly, our preliminary data also suggests that as the number of women represented in the magazine increases, the proportion of women in advertisements decreases. Our poster will focus on a close examination of the data and sociopolitical context of 1965-1990 in order to fully explore this potential correlation. We will also discuss our methodology and include a few examples of our visualizations.

This project aims, not only to gain insights from an analysis of
Time magazine and to make these insights publicly accessible, but also to establish novel methodologies for the visual analytics of large data sets, particularly of image-based corpuses, which we hope to use for years to come and to share with other researchers.

The ultimate goal of this project is to create a website with contextualized interactive visualizations based on the entire archive. Our initial approach was inspired by Manovich’s Selfie-city and Photo-trails work, and by his team’s use of direct visualization (Crockett, 2016), which we see as a way to engage broad audiences into complex corpuses. We also draw inspiration from
Robots Reading Vogue (King and Leonard) and
Neural Neighbors (Leonard), which are projects based in the Yale University library system. By exploring specific, humanities-based research questions in this early phase of our project we will be able to make meaning and better contextualize the interactive visualizations in the end


Note:
Will there be a bibliography for these citations.

.