Abstract
Described method uses normalisation of lexical frequencies taking into account dispersion and corpus composition. Statistical significance measures are finally used for evaluation of frequency differences that can characterise usage changes.