Evaluating the Potential of a Large-Scale Polysemy Network as a Model of Plausible Semantic Shifts

We present a very large network of crosslinguistic polysemies, and compare the notion of semantic relatedness it encodes to the catalogue of semantic shifts maintained by the Russian Academy of Sciences. We separately evaluate all types of semantic shifts featured in the catalogue, including shifts occurring during semantic evolution, during borrowing, and during morphological derivation. The comparison shows that over one third of the attested semantic shifts take place between close neighbors in the network. This can be considered strong evidence for the usefulness of polysemy networks in modelling most types of lexical change, making them a valuable resource e.g. for semantic reconstruction or future automatization of cognate detection. We also show that the semantic shifts which occur during morphological derivation form a divergent class, and might need to be modelled separately