On two mathematical representations for “semantic maps”

Abstract We describe two mathematical representations for what have come to be called “semantic maps”, that is, of typological universals linguistic co-expression with the aim inferring similarity relations between concepts from those universals. The are a graph structure and Euclidean space, latter as inferred through multidimensional scaling. Graph in types. In both types, meanings represented vertices (nodes) edges (links). One representation is pairwise graph, which represents all graph; an example CLICS. other minimally connected – “classic semantic map”. This only necessary maintain connectivity, principle expressed by single form make up subgraph whole graph. space points, distance specified number spatial dimensions. focus on proper interpretation types representations, algorithms constructing measuring goodness fit data, balancing informativeness representation.