This project explores the intricate ways contemporary narratives of migration—an inherently traumatic journey—represent and/or work through the migrants’ trauma, utilizing a blend of digital humanities tools and traditional literary analysis. Through the implementation of a mixed methodology entrenched in postcolonial studies, trauma theory, computational narratology, and sentiment analysis, this study aims to investigate how the frequently over-applied label of trauma (Arnolde-de Simine 2018: 151-52) may be incorporated into the broader context of novels featuring migrant characters. Employing a mixed methodology of distant and close reading, rooted in digital and traditional literary studies, allows for highlighting trauma writing as an abundant and predominant mode of migration fiction while showcasing the intricacies and uniqueness of each narrative.
Departing from trauma theory, this project is based on interpreting trauma as a multifaceted phenomenon that varies contextually, resonating with postcolonial perspectives on diverse migration narratives. Trauma theory, particularly as initially conceptualized by seminal scholars like Cathy Caruth (1996), has been subjected to criticism for its tendencies to depoliticize and universalize the traumatic experience (Craps 2013; Rodi-Risberg 2018). Nevertheless, trauma theory may serve as an optimal theoretical framework for a computational model of affect in migration narratives because it provides multiple possibilities for detecting and interpreting ‘symptoms.’ A computational reading results in a metaphorical reading of narrative patterns as symptoms and interpreting the characters’ emotional arcs as their journey through traumatic memories.
In the initial phase of the project, trauma featured in a small corpus of migration novels is looked at through the lens of witnessing, specifically what Stef Craps (2013) has termed ‘postcolonial witnessing,’ which dwells on marginalized narratives and recognizes their specific contexts and nuances while decentering the dominant Eurocentric perspective. Moreover, the project moves into finding means to translate the most common tropes for representing trauma in fiction, namely “absence, indirection, and repetition” (Pederson 2018), into a computational concept. However, in order to respect the nuances of these literary narratives and avoid reducing lives to numbers, a mixed methodology is adopted to highlight a more ethical approach towards the corpus.
The mixed methodology approach combines distant and close reading in an iterative process; an approach that has been called “computational hermeneutics” (Piper 2015), “parallax reading” (Sample 2017), and “middle reading” (Elkins and Chun 2019), among other identifiers. Distant reading here comprises sentiment analysis, machine learning for pattern recognition, and text mining for contextual analysis to provide insights into how these themes are integrated into the narrative structure. These distant reading methods combined with close reading allow for computational hermeneutics and highlight the multifaceted ways of representing trauma through fiction.
Due to its limitations when it comes to literary settings, sentiment analysis in the context of narrative and literary studies has received less attention and seen smaller success (E. Kim and Klinger 2018; H. Kim 2022; Rebora 2023). Therefore, sentiment analysis here is adapted to a novel setting to overcome its typical quantitative limitations. In order to circumvent this problem, the project employs SentimentArcs, the “largest ensemble of open NLP sentiment analysis models” which allow for an approach Jon Chun calls “self-supervised machine learning” (Chun 2021).
By adopting a methodology similar to Elkins (2022), the project oscillates between testing these models on the corpus and closely examining the passages highlighted through the model. This adaptation is a crucial step for capturing the complex emotional landscapes of trauma in literature. This involves a detailed examination of literary devices, narrative techniques, and character development to interpret the emotional nuances and thematic complexities identified through digital analysis. Close reading offers a detailed, interpretive analysis of texts, considering literary devices and narrative structures and addresses the challenges of applying sentiment analysis to literary studies.
The initial analysis using sentiment analysis and text mining has uncovered varying emotional landscapes in these narratives, reflecting the complexities of trauma in migration experiences. Patterns of sentiment shifts correlated with key narrative events have been observed. These preliminary findings demonstrate not only the varied representations of trauma but also the potential of this hybrid approach to yield deeper insights into the narratives. The close reading phase is further enriching these findings, providing a nuanced understanding of how these themes are woven into the literary fabric of the novels.