By Vincent Labatut
My name is Vincent Labatut, from Avignon Université, France. This academic year, I am a visiting scholar at MACSI (Mathematics Applications Consortium for Science and Industry) at the University of Limerick. I am a computer scientist specializing in data mining, text mining, natural language processing, and network science. In the latter field, my recent work focuses on the extraction and analysis of character networks from works of fiction (TV series, novels, comics, movies, etc.) [1]. These are graphs in which vertices represent characters, and edges model the relationships or interactions between them. A few years ago, I began collaborating with Pádraig MacCarron from MACSI, which led to some Irish students traveling to Avignon in order to work directly with us, followed by my visit to Limerick as a visiting scholar to deepen this collaboration.
One of our articles recently published [2] explores the alignment of character networks across different adaptations of the same narrative, providing insight into how the same story is transposed from one medium to another. We focused on the case of G. R. R. Martin’s “A Song of Ice and Fire”, comparing the original novels to their comic book and TV series adaptations (i.e. HBO’s “Game of Thrones”). The main challenge in this study is matching characters between the different versions of the story, as the adaptation often involves removing, creating, or even splitting and merging characters. Our key finding demonstrates that analyzing character interactions alone is insufficient for this task, even when considering the dynamics of the story. Instead, it is essential to incorporate a minimal amount of textual content (in our case, summaries of chapters, issues, and episodes). Identifying character correspondences enables the comparison of the fundamental building blocks of the narratives. As an illustration, the figure below shows a comparison of each pair of narratives: correct matches are shown in green, while incorrect ones appear in yellow (false negatives) or purple (false positives). When comparing the TV series to the novels, the moment when the adaptation begins to diverge significantly from the original material is clearly visible in the plot.
Figure 1 Best performing alignment for all pairs of media, from top to bottom: a) Novels vs. TV Show, b) Novels vs. Comics, and c) Novels vs. Comics. Green denotes true positives, red false positives, yellow false negatives and purple true negatives
My stay in Limerick provides an excellent opportunity to continue working with Pádraig MacCarron, facilitating more direct, interactive discussions that enhance both productivity and creativity. Our current work research examines the impact of missing sources on the topology of character networks extracted from mythological material, using superhero comic books as a modern (and readily available) proxy for Irish mythological texts [3]. This work will be presented at the upcoming NetSci conference.
Overall, my experience at the University of Limerick has been outstanding. The people are incredibly welcoming, the institution offers all the necessary resources for research, and the campus is very beautiful. In the mathematics department, I have met a group of highly talented and dynamic researchers, I have been exposed to research themes that are new to me and have broadened my perspective of the field, and I initiated new collaborations. Additionally, I have begun working with the social dynamics group in the psychology department, which explores social networks, among other topics. I highly recommend the university, particularly MACSI, for anyone considering a similar research stay.
References
- V. Labatut and X. Bost, “Extraction and Analysis of Fictional Character Networks: A Survey,” ACM Computing Surveys, vol. 52, no. 5, p. 89, 2019. DOI: 10.1145/3344548
- A. Amalvy, M. Janickyj, S. Mannion, P. MacCarron, and V. Labatut, “Interconnected Kingdoms: Comparing ‘A Song of Ice and Fire ́Crossmedia Adaptations Using Complex Networks,” Social Network Analysis and Mining, vol. 14, p. 199, 2024. DOI: 10.1007/s13278-024-01365-z
- P. MacCarron, V. Labatut, and J. Garcke, “Impact of Incomplete Data on Character Network Extraction: The Case of Comics,” in NetSci Conference, 2025.