Abstract
By employing the grammatical profile method, we show to what extent patterns of use of grammatical features are interpretable through word meaning and how stable this relationship is across different corpora. we extracted grammatical profiles of all nouns in four different corpora of Czech (modern written Czech, spoken Czech, Czech movie subtitles, Czech translations of the proceedings of the European Parliament). After filtering out low frequency lemmas (ipm < 7), we analyzed a set of 821 noun lemmas that appear in all four corpora.
The distributions in each corpus were analyzed using hierarchical clustering to identify lemmas with similar profiles. Obtained clusters were studied with respect to the relative frequency of individual cases and the meanings of cluster members.
The results show that there are potentially meaningful groupings in the data. We identified four clusters that are, with one exception, found in all four corpora.
The clusters suggest that the analyzed distributional patterns may be interpreted as an interplay between word meaning and more abstract syntactic patterns.