> is this basically just running billions of text passwords through the md5 algorithm and comparing the results with the unknown string?
Yes: "using a computer that tried more than 300 billion plaintext guesses every second"
The Arabic version was obtained by doing an educated guess based on the English version