Slashdot Mirror

The entropy and other statistical measures of the Voynich language is different from Indo-European languages. Zandbergen goes through this in some detail. To quote

Voynichese is nearly as information-rich as Julius Caesar's Latin, and significantly more so than the Vulgate version of Genesis.

Voynichese is less information-rich than Latin in the first two characters of each word, but compensates by greater variability in the trailer.

and

The statistics of Voynichese and a Mandarin text written in the Pinyin script (using a trailing numerical character to indicate tone) are very different.

There is actually a lot more of this in this and other papers. The Voynich language, for another example, has a lot more repeated words than (say) English. I seem to remember that the closest match in terms of word repetitions was with Vietnamese, and there was some speculation that it might be an invented script for that language, but that didn't pan out in detailed examination. The upshot is that it is just not realistic to just assume that Voynich is a common language written in some weird script (and, also, that these substitution games have been played before).

Send them the Voynich Manuscript.

That should buy some time.

something that can make sense of the voynich manuscript http://www.voynich.nu/. They should have tested their system on it.

Like this?

Not quite your scenario, but the likeliest explanation for that document is basically what you outline, and the result is exactly what you anticipated. A lot of people have gotten very excited about it, but the simplest explanation that fits the facts is that somebody just faked the whole thing for kicks.

The letter entropy in the Voynich MS is quite different from Latin or English, with the first character having less entropy, the second a lot less. and the third and fourth a lot more entropy (more entropy means less predictability).

I do not see how grill gibberesh could easily do this, and note that the Nature paper explicitly did not consider letter entropy.

it seems to me that where the burden lies is not so clear. If I found a 4+ century old manuscript in an unknown language, I would assume it to be real, unless proven otherwise.

Here is what is known about the manuscript

It is not a moden hoax, as old letters have been found after its rediscovery that refer to it.

It has 234 pages (plenty long enough for statistical analysis) and appears to have been copied by a professional. It also has a number of images of various sorts, with "labels" and the words used in the labels also appear in the text near the images.

(BTW, I do not see how you get that in a "grill" system hoax.)

There are, however, no apparent images of the usual alchemical signs, occult signs, etc. - the sort of stuff that might impress an occult minded royal buyer.

It appears to have been written in (at least) two langauges or dialects or jargons, based on word use, with each page being in only one "language"

Here is a plot showing the correlation of word usage between pages in the manuscript - color coded with red meaning more words in common, black meaning the fewest, with page one in the upper left hand corner.

I would expect a grill method to produce a random version of this image - which is clearly not random.

The text follows roughly the 1st. and 2nd. Zipf's laws of word frequencies.

The word length distribution is very different from Latin, German, English, French or Italian, and, in fact, is similar to various Asian languages like Chinese - words are uniformly short.

Again, I don't see why a grill method would do this.

The 2nd. order entropy is too low for an European language using a simple substitution cipher, and the third and fourth are too high. (Also, download this pdf.)

What does this mean ? It means that the second character of a word in the manuscript is more predictable than in a typical European language, and that the third and fourth characters are _less_ predictable.

It is very hard for me to see how this could come from the grill method.

So, regardless of where the burden of proof lies, I, for one, am not convinced.

it seems to me that where the burden lies is not so clear. If I found a 4+ century old manuscript in an unknown language, I would assume it to be real, unless proven otherwise.

Here is what is known about the manuscript

It is not a moden hoax, as old letters have been found after its rediscovery that refer to it.

It has 234 pages (plenty long enough for statistical analysis) and appears to have been copied by a professional. It also has a number of images of various sorts, with "labels" and the words used in the labels also appear in the text near the images.

(BTW, I do not see how you get that in a "grill" system hoax.)

There are, however, no apparent images of the usual alchemical signs, occult signs, etc. - the sort of stuff that might impress an occult minded royal buyer.

It appears to have been written in (at least) two langauges or dialects or jargons, based on word use, with each page being in only one "language"

Here is a plot showing the correlation of word usage between pages in the manuscript - color coded with red meaning more words in common, black meaning the fewest, with page one in the upper left hand corner.

I would expect a grill method to produce a random version of this image - which is clearly not random.

The text follows roughly the 1st. and 2nd. Zipf's laws of word frequencies.

The word length distribution is very different from Latin, German, English, French or Italian, and, in fact, is similar to various Asian languages like Chinese - words are uniformly short.

Again, I don't see why a grill method would do this.

The 2nd. order entropy is too low for an European language using a simple substitution cipher, and the third and fourth are too high. (Also, download this pdf.)

What does this mean ? It means that the second character of a word in the manuscript is more predictable than in a typical European language, and that the third and fourth characters are _less_ predictable.

It is very hard for me to see how this could come from the grill method.

So, regardless of where the burden of proof lies, I, for one, am not convinced.

In case you're wondering what it looks like

http://www.voynich.nu/