What kind of corpus is a web corpus?

This paper discusses an investigation into the Norwegian NoWaC corpus. We have com-pared this web corpus with one corpus of spo-ken language and one of written language. For nearly all variables that we look at, the web corpus sides with the written corpus, not the spoken one. Thus, despite including lan-guage samples from blogs and web forums, NoWaC does not appear to be more speech-like. One exception is interjections, which it does have to some larger extent than the writ-ten corpus. It also has taboo words, lacking in the other two. The comparisons have been made purely on the basis of frequency lists, showing that this is a possible and simple way of comparing corpora. We use both a qualita-tive and quantitative method. In the latter, (log) relative frequency plots show an almost linear relation between NoWaC and the writ-ten corpus.