




Can computational models capture semantic shift in the concept of the HIV/AIDS pandemic and its associations with race and sexuality? This study combines historical and computational linguistic methods to study American public discourse during the HIV/AIDS pandemic between 1980 and 2010. We developed a conceptual model of HIV/AIDS based on historical literature, in which we hypothesized that HIV/AIDS was strongly associated with gay men and drug users during the early 80’s, which slowly moved to being more associated with Sub-Saharan Africa. This hypothesis was then tested with two experimental techniques using diachronic word embedding models. Our results suggest that distributional semantic models showcase the expected conceptual changes, depending on the type of model. Overall, this research shows that interdisciplinary approach using computational tools in combination with historical research enables research of complex social issues.