Culture, corpora and semantics is a methodological investigation in the use of elicited data and Web data in the analysis of cultural specificities starting from semantic elements. After considering and discussing several theoretical and analytical approaches to culture in linguistics, anthropology, psychology, and marketing research, a specifically developed method of analysis and cross-cultural comparison is applied to elicited data on chocolate and wine, gathered through free sentence-completion and sentence-writing tests on English and Italian respondents. The results obtained are discussed within the framework of cultural systems theories and used as control reference for further methodological investigations. In particular, the elicited data are qualitatively and quantitatively compared to non-elicited sentences on chocolate and wine from general Web corpora in English and Italian. Furthermore, in order to find an alternative route which could avoid the complex and time-consuming process of manually coding a large dataset, some alternative routes are tested, based on the creation of sub-corpora using sampling procedures and analysis of a limited number of the most frequent words in the dataset’s wordlist. Finally, an automatic semantic tagger is tested on the elicited data, in order to assess the extent of its possible application in cultural analysis. Comparisons between the Web corpora and the elicited data suggest that large general Web corpora can be considered representative of the cultural associations to a node word and could thus be used in cultural analysis or in exploratory marketing research. Finally, in the light of the results of the various methodological tests, the work discusses general issues, such as the relationship between word frequency and cultural relevance, and tagset granularity.