Analyzing suicide life stories on Wikipedia with Highway_star and other textual visualization tools.

Being responsible for a death every 40s, suicide is a major public health concern (Brunier et al. 2019). Even if many of its risk factors are social (Van Orden et al. 2010), there are surprisingly few qualitative sociological studies about the phenomenon. This study aim is to provide a life-story sociological analysis of suicidal trajectories. Two challenges are identified: gathering suicidal narrative and maintaining a quantitative foreground in order to deepen and rationalize the interpretation of data. They are both faced using a self-made, free to use, open access, algorithm: Highway_star (https://github.com/matheo-daly/highway_star). Two corpora of Wikipedia biographies of people who died by suicide in the 1920s (N = 82) and 2020s (N = 49) are gathered. Following an application of Fritze Schütze's methodology (Schütze, 2014), classical textual visualizations are produced. A Hierarchical Descending Classification, a Factorial Correspondence Analysis and a Similarity Analysis reveal five narration categories centered around different topics: cinema, death, family, poetry and politics. As none of those visualizations focuses on the developmental aspect of the biography, they offer limited interest for a life-story investigation. The second functionality of the Highway_star tool, which represents a narrative's unfolding with a Sankey Diagram, allows completing the analysis. It shows interesting differences between decades or gender. An example of the last being that men narratives tend to be more complex and achievement focused, while the women ones are more linear and family centered. The study's range has limitations. A major one is related to the corpus and the inability to identify clearly which parts of the narratives are associated to fame and which to suicide. Another one is linked to the Highway_star tool that sometimes lack of flexibility.