The present study pertains to the field of oenological discourse, which is defined by its focus on the description of wine production and tasting represented by a variety of text types and stylistic genres. Drawing on a representative corpus of wine descriptions, the authors seek to reveal the typical language used to specify the organoleptic properties of wine, such as aroma and flavour. Due to the lack of specialized items denoting subtleties of smell and taste, metaphors are lavishly used, mostly synaesthetic ones. Ontological metaphors are also present, mapping qualities of humans or inanimate objects on wine. Apart from metaphors, association-based references to various fruit, berries, and flowers are quite common. All the metaphors and associations occur in a limited set of grammatical construction types. The study explores correlations between particular grammatical constructions, metaphor types, and thematical sections of wine descriptions (aroma, flavour, aftertaste). To this purpose, two computational techniques have been applied: distributional semantics and cluster analysis. The former has shown that associations play a central role in rendering wine aroma, synaesthetic metaphors are instrumental in specifying both aroma and flavour, while ontological metaphors are occasionally used to characterize flavour and aftertaste. Metaphor types and grammatical constructions are shown to be inter-related. Besides, aroma and flavour descriptions often happen to be dependent on wine colour (red, white, rose). The application of cluster analysis has corroborated the above-mentioned link between metaphor types and grammatical constructions. Some clusters also tend to be thematically homogeneous, reflecting a particular aspect of perception, e.g. aftertaste or wine body. Thus, the methods applied have yielded compatible results. The study demonstrates the validity of computational techniques in analyzing thematic text varieties and opens up the prospects of their automatic generation.