Figure 3: (A–D) TFMN frames around “woman” (A), “man” (B) and “person” (D). Words are clustered in communities obtained via the Louvain algorithm (cf. Konstantinidis, Papadopoulos Kompatsiaris, 2017). Words in the same community of “woman” in A are plotted through a hierarchical edge-bundling visualisation in C. Positive (negative) words and links are highlighted in cyan (red). Links between positive and negative concepts are reported in purple. Semantic links between synonyms are in green. (E–F) Emotional profiles for the frames of “woman” (E) and “man” (F) indicate the fraction of concepts in a frame eliciting a certain emotion.