Project Overview

This dashboard is the final project for Spring 2023’s ANTH630: Statistics for Anthropology at the University of Maryland taught by Madeline Brown. The project was completed by Britney Bibeault and Ia Bull, students of the PhD in Information Studies program. Quantitative approaches are fairly common in translation studies and lexology as they are in literary studies, though they both often have their niches and blind-spots (Guo 2022). We hope to broadly contribute to the growing trend of interdisciplinary studies while showing how useful archival data can be when reused.

Data Source

This dashboard highlights the literary translation dataset collected and written about by Erlin et al. (2022) in “The TRANSCOMP Dataset of Literary Translations from 120 Languages and a Parallel Collection of English-Language Originals,” set of over 10,000 literary works that were translated into English in the early 1900s and English-origin works of the same time-period. This data was pulled from a larger dataset (hereafter referred to as NovelTM) of nearly 200,000 literary works located in the HathiTrust Digital Library published since the eighteenth century (Underwood: 2020). We should address the limitations in this source of data. The HathiTrust digital library is not a comprehensive representation of past literature. It is primarily composed of books from academic libraries in the United States, resulting in a bias towards famous authors and weaker coverage of popular culture and juvenile fiction, particularly in non-Anglo-American contexts. Furthermore, the collection is incomplete, containing only slightly more than half of the 19th century fiction titles mentioned in Publishers Weekly and less than a quarter of 20th century titles (Underwood: 200, 4-5).

The dataset we are using can be found on the Journal of Open Humanities Data [add link here].

Research Questions

By applying statistics to this dataset, we hope to answer (1) What are the general trends in this dataset in relation to what languages were translated? (2) Is there a correlation between time and languages in terms of when different languages were translated?

Expected Impacts

One of the main impacts we expect of this project is to bring more awareness to the reuse of open data, especially of humanities data, to encourage other researchers to submit their data to open access resources and to do subsequent analyses on data so more questions can be answered without needing to collect new data. this research project is also shedding a spot-light on previously unexplored areas of research identified by Erlin et al. to encourage others to engage this data critically and more thoroughly. They identified two areas of research that could benefit from access to their data, which include investigating the structural asymmetries in the global flow of translations and linking these asymmetries to differences in the linguistic, stylistical, or thematic features of translations. However, the current project cannot engage directly with these specific areas due to the scope and limitations of the research. Nonetheless, the project’s findings can still contribute to the broader field of translation studies and provide valuable insights into the patterns and characteristics of translated texts in the global context. By building on the work of Erlin et al. and other scholars in the field, this research project can help expand our understanding of how translations reflect and shape cultural biases and power dynamics in the “world republic of letters” (2022), and in what ways the written record overshadows other forms of influence (Donald 2023).

Overview

The dataset used in this dashboard is the compiled metadata derived from the “TRANSCOMP Dataset of Literary Translations from 120 Languages and a Parallel Collection of English-Language Originals” by Erlin et al. (2022). The dataset consists of over 10,000 literary works translated into English in 1900s, along with English-origin works from the same time period. The data was extracted from the larger NovelTM dataset, which contains nearly 200,000 literary works published since the eighteenth century and is located in the HathiTrust Digital Library (Underwood et al. 2020).

It is important to note the limitations of the data source. The HathiTrust digital library primarily includes books from academic libraries in the United States, resulting in a bias towards well-known authors and a weaker representation of popular culture and juvenile fiction, especially outside Anglo-American contexts. Additionally, the collection is incomplete, covering only a fraction of the titles mentioned in Publishers Weekly for the 19th and 20th centuries.

Through the analyses we ran on this data, we found the following findings interesting and warranting future research; 1) texts published in different languages experienced different amounts of attention by English translators. The historical context around which those spikes or drops occur could allow for more insight regarding the trends in translations. 2) There are a number of different opportunities to evaluate the state of literary works over time through the TransComp dataset collected by Underwood et al. (2020) and developed by Erlin et al. (2022). One visualization that we never engaged in was visualizing the global presence of the languages being translated into English. One way to do this is with a heatmap imposed on a global Choropleth in leaflet, in combination with or instead using R packages Lingtypology (Moroz 2017) or glottospace (Norder et al.).

Research Question 1 focuses on the top translations, including English, in the dataset. A bar chart displays the counts of translations for each language, with English having the highest count due to the dataset’s focus on translations into English. Research Question 2 further explores translations by year and language. A line chart shows the number of translations over time, allowing for comparisons between languages. The dashboard also presents a plot of translations over time, showing fluctuations in the number of translations into English.

Overall, this dashboard provides an initial exploration of the TRANSCOMP dataset and offers avenues for further investigation into translation trends and their historical context. This work highlights the need for examining historical contexts to gain deeper insights into translation trends and emphasizes the potential for comparing this dataset with modern translation data.

Publication

It is our plan to write a paper on the process of creating this project to submit to the Journal of Open Humanities Data in the coming year to share how we reused and applied statistics methods to humanities data, in hopes of encouraging more creative uses of existing datasets.