build a bpe on wikipedia

[jwijffels]

We can use e.g. data at https://linguatools.org/tools/corpora/wikipedia-monolingual-corpora/
Steps:

• For each xml file in wikipedia
• read_xml("C:/Users/Jan/Desktop/sample-monolingual.xml") %>% xml_text()
• Append that text to 1 file
• Build the model
• Save the model
  Looks pretty easy.
• Probably run this on a more beefy machine