I'd rather say the corpus itself. Of course scrapers might play a role in creating that corpus. But it’s just one part: maybe also OCR and OCR correction, forma

I'd rather say the corpus itself. Of course scrapers might play a role in creating that corpus. But it’s just one part: maybe also OCR and OCR correction, format conversion (think of double column PDFs).