Leveraging Computational Methods for Comparing Low-Prestige Translations

Despite decades of scholarly interest in computational literary studies, modern children’s literature has largely escaped notice by digital humanities scholars. A 2020 conference at the University of Antwerp noted that “Digital Humanities has had a big impact on the field of literary studies as a whole, but its presence in children’s literature studies has been limited so far.” The copyright considerations involved in working with modern texts and the low prestige associated with children’s literature provide both legal and social disincentives for working with these texts. Nonetheless, some American mass-market “series books”-- which were decried by children’s librarians for their lack of literary merit (Greenlee et al. 1995)-- found audiences across the globe through long-running translation efforts in many different countries. These translations have the potential to shed light on topics ranging from the expansion of multinational corporations to the growing familiarity of American cultural practices.

This poster will demonstrate the results of applying computational methods to compare the original English volumes of The Baby-Sitters Club, a series of over 200 books written by Ann M. Martin between 1986 and 2000, with translations into six languages: French (translated three times, in Quebec, Belgium, and France), Dutch, German, Peninsular Spanish, Italian, and Russian. These translations represent a range of time periods (from the French and Belgian translations starting in 1990, to Russian, which was first published in 2020 in conjunction with the release of a new Netflix series available in many languages) and levels of success (fewer than five books were translated into Russian, vs. over 60 in Quebec and the Netherlands). These translations were produced quickly, under a demanding monthly publication schedule. The lack of attention to detail is part of what makes them remarkable; earlier work by part of this group has identified inconsistencies even on the level of an individual book. While there has been previous work applying close reading to individual translations of the Baby-Sitters Club books (Desmet 2007, Aisyah 2013), to date there has been no scholarship that has taken such a broadly comparative approach, let alone with digital methods.

The focus of our comparison will be centered on food terminology and character descriptions in order to determine the “relatability” of the narrative. The original Baby-Sitters Club books make extensive use of very culturally specific common food items, like peanut butter and jelly sandwiches and M&Ms candy, which tend to be translated directly (leading to a sense of estrangement) or elided (leading to a genericization of the US-specific context of the series). Yet, even if the setting of the series is established to be unfamiliar to the reader, as the result of a decision to not localize the text, there is the possibility of the reader connecting with the characters, who form a club of 7 distinct teenage girls. Some of the criticism of this series relates to its formulaic nature, which includes a reliable “Chapter 2” consisting almost entirely of a description of how the Baby-Sitters Club works, along with each individual member. For this poster, we will, through close reading, evaluate how relatable the characters are, based on the specific word choices made in each language.

For this poster, we used WordNet with manual correction on the original English text of a subset of the overall Baby-Sitters Club corpus to identify sentences that include food-related terms. We have used word lists and an English SpaCy NER model to identify original sentences with food, and used the Bleualign algorithm for sentence alignment between the original English and the translations, to flag translated passages for close-reading and evaluation. This approach has surfaced noteworthy differences in food translation strategies. For instance, M&Ms were replaced with Bonitos in the Belgian translation. When that translation was later adapted in France, the re-insertion of M&Ms was one of a small number of significant changes. Some translation strategies are specific to a given translator, such as Dominique Laplier’s tendency to replace an American brand with a phonologically-similar French one, even when that brand is a very different food (e.g. ‘Doritos’ corn chips to ‘Bonitos’ chocolates).

This project can contribute to the broader field of DH by offering annotated children’s literature as a ground truth source for NLP (in a manner compatible with current copyright law in the US, see Bamman et al. 2019), by serving as a case study for the use of DH in translation studies, and as a step towards further research on the effects of localization in the global flow of popular culture.