An approach to cross-concept cognacy identification

Abstract

It is a well known phenomenon in historical linguistics, that the meaning of a proto form is different to the meaning of its descendants. This phenomenon of meaning change is often ignored in studies which use tools from statistical phylogenetic analysis to determine language relationships. It has been shown, that the databases currently used in linguistic phylogeny exhibit a considerable amount of the described phenomenon. The current study proposes a method to detect such instances of cross-concept relationships of words. Although the evaluation can not be done by standard means, the results indicate that semantic similarity is a good indicator for cross-concept relationships and that tools from computational biology offer a good framework for this kind of approach.