Abstract
The paper presents a corpus-driven method for the detection of recent grammatical change in contemporary Czech newspapers. It is based on a large and homogeneous material (825 million tokens of a single newspaper) that covers a 23-year time span.
The task is operationalised into finding the most relevant frequency change manifested by selected subsets of the Czech tagset. The results show changing proportions of parts of speech, nominal cases etc. that indicate a shift towards more "verbal" language associated with increasing informality of the newspaper register.