Stellar, what Anton wants to know is the following: if I take any paragraph from the VM and present it to five people to decipher using your method, will the results be the same? I think the answer is no, unless you tell them beforehand what their translation should look like.
In other words, the method you propose requires one to know the plain text, rendering the encription useless.
Additionally, the translation is a version with deviant syntax, spelling, word order and vocabulary. So even if you gave people the normal version of the text that was encrypted, they will still come up with different translations because there's a lot of fidgeting and guesswork required of the translator.
So what Anton means is that if we assume repeatability as a requirement for a solution - which we should - then your method does not survive the test. It's mostly guesswork and everybody will guess differently unless you tell them what to guess.
In other words, the method you propose requires one to know the plain text, rendering the encription useless.
Additionally, the translation is a version with deviant syntax, spelling, word order and vocabulary. So even if you gave people the normal version of the text that was encrypted, they will still come up with different translations because there's a lot of fidgeting and guesswork required of the translator.
So what Anton means is that if we assume repeatability as a requirement for a solution - which we should - then your method does not survive the test. It's mostly guesswork and everybody will guess differently unless you tell them what to guess.