The use of English, Czech and French punctuation marks in reference, parallel and comparable web corpora: a question of methodology

This paper analyses the frequency of six punctuation marks (the comma, period, colon, semicolon, question mark and exclamation mark) in three languages (English, French and Czech) in three different types of corpora — comparable web corpora, large monolingual general (reference) corpora and parallel (translation) corpora. The aim of the analysis is to identify which type of corpus and which methodology are the most suitable for contrastive research into punctuation. The data shows that the frequency of different punctuation marks is very sensitive to the text type. Therefore, the web corpora, containing uncontrollable amounts of various text types, cannot provide specific and reliable information about the use of punctuation marks in a given language. We argue that despite their limitations in terms of size and composition as well as the potential specific features of the language of translation, the parallel corpora used in combination with the general (reference) corpora provide the best data for such research.