Faculty Sponsor: Typhaine Leservot
Lucas Donat
Lucas is a rising senior (’26) from Fairfax, Virginia studying French and Data Analysis. His interests include postcolonial literary theory, translation, and abolition geography. At Wesleyan, he has worked as a course assistant and tutor for the French department. In his free time, Lucas enjoys making music with friends, playing volleyball, and reading.
Abstract: In postcolonial literature (literature written by authors from colonized and/or formerly colonized territories), artists have long created art that revisits past colonial violence inflicted on their home nations by the colonizers. For authors from North Africa living in North Africa and France from the 1950s until the end of the 20th century, this has meant focusing on revealing living conditions under France’s subjugation. Analyzing how this relationship has evolved in the 21st century in literature written by authors from North Africa now living in Québec, geographically separated from both North Africa and France, can lead to deeper understandings of postcolonial thought and its recent evolutions. Also in recent years, developments in natural language processing (NLP) have led to the creation of text embedding models, which capture more nuance than previous computational text analysis techniques. This project aims to explore ways in which text embedding analysis can provide fruitful information for literary analysis through grouping processes that can reveal unseen semantic similarities. Principal Component Analysis (PCA) and Constrained K-Means clustering on 101 quotes mentioning France from 8 novels written by two Algerian authors living in Québec finds that 2 clusters of sizes 50 and 51 was the most viable grouping arrangement across 10,000 simulated trials (n = 1766, 17.66%). Chi-square difference test showed a significant relationship between author and cluster membership (p = 0.00886) and novel and cluster membership (p = 0.01255), though no absolute groupings were found in either case. Further, passages in cluster 1 tended to negatively mention France in the past tense, both in political and personal contexts, while texts in cluster 2 tended to more neutrally mention France in contexts of possible future scenarios. These tendencies were consistent regardless of author or parent text.
Donat-QAC-Apprenticeship-Poster