The NIKAW project (Networks of Ideas and Knowledge in the Ancient World) investigates the flow of knowledge in Ancient Greek and Latin literature using Natural Language Processing (NLP) and Social Network Analysis (SNA). This presentation will provide an overview of the project’s progress, with an emphasis on the development and population of our multi-modal graph database, which captures mentions of individuals within these texts.
The core of the discussion will focus on the interpretability and significance of two potentialapproaches for modeling co-occurrence of mentions derived from this multi-modal graph. The first approach employs co-occurrence-based edges, where connections are either binary, indicating the presence of individuals within the same text unit, or weighted, reflecting the distance (number of tokens) between mentioned individuals. The second approach utilizes similarity-based edges, established based on statistical similarities, such as mention embeddings or entity profile vectors that represent the frequency or weighted count of co-occurrences with other entities.