As explained in Section 4.1 of Roland Schäfer (2015) Processing and querying large web corpora with the COW14 architecture, we have to take certain measures in order to stay within the bounds of German copyright laws. This means that we only release sentence shuffles, i.e., corpora which are just bags of sentences. In other words, there are no documents in released versions of COW corpora, just single sentences without contexts. The original URL plus some other meta data are recorded for each sentence, however.
A web-derived corpus crawled in the DE top-level domain in late 2011. Created with texrex-mrvain (deprecated) and TreeTagger. The DECOW12Q is a subset that contains only documents written in a quasi-spontaneous register, selected based on the occurrences of cliticized variants of the indefinite article as described in: Schäfer and Sayatz (2014) Die Kurzformen des Indefinitartikels im Deutschen. Zeitschrift für Sprachwissenschaft 33(2). Continue reading