[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: VMs: algorithm to generate VMS like text

There seems to me to be a problem with a program that produces VM like text that is only tested against entropic measurements-that is that there must be a great number of schemes that will give the same statistics, but may not have any other similarity to VM text.   In other words, the inversions are not unique; it might look like a duck, walk like a duck, and even quack, but can you eat it? A test metric is needed that might include entropy, but should have other tests as well, IMHO.
-----Original Message-----
From: owner-vms-list@xxxxxxxxxxx [mailto:owner-vms-list@xxxxxxxxxxx]On Behalf Of Jeff
Sent: Thursday, August 14, 2003 12:16 PM
To: vms-list@xxxxxxxxxxx
Subject: Re: VMs: algorithm to generate VMS like text

>> Re: VMs: algorithm to generate VMS like text
>> Hi Rene I'll have a look at producing Courier and EVA. The algorithm uses data from a >> >> previous program that analysed all of the FSG text in the interlinear 1.7 file and produced a >> list of contact frequencies for each "letter". This is then plugged in to the text generator as a >> look up table for each letter. I then seed a random number generator and based on the
>> current letter, do a lookup on the table of right hand contact frequencies for that letter to
>> produce the next letter in the sequence. I have treated space as a letter here. So, to answer
>> your "blue sky" question, there is no input text and the output is different each time the
>> program is run but still has (or should have) all the characteristics of FSG text. proviso here
>> is - the look up tables for each letter are tables of 100 letters, so any letter that contacts
>> less than 0.5% has been dropped I then thought that if an algorithm can produce text that is
>> alike to the VMS and cannot be proved statistically to differ from the VMS then that
>> algorithm could be a computational representation of a physical process used to generate
>> the text at random. I have an idea of how this might work but want to spend a bit more time
>> on it before posting to the list. Anyway, I'll change the program to produce Courier and EVA
>> first and get some stats from Monkey.
>> Regards,
>> Brett

> Rene Zandbergen r_zandbergen@xxxxxxxxx wrote:
> Brett Cotton wrote:
> a week or two ago a there was a post saying that it would be difficult to generate VMS text
> using a an algorithm. I brought together some thoughts that I had recently and now have a
> (not quite perfect yet) program that will generate any amount of VMS style text. Here is the
> output of the program when generating 15 lines of text: Input number of lines
> That looks quite interesting indeed. Most people (including myself) are not too familiar with
> the FSG alphabet, so it is hard to judge. In fact, this could be a sample taken straight out of
> the VMs and I wouldn't be able to say :-) The check with Monkey should be quite interesting,
> and should of course be compared with samples also done in FSG, unless you make the
> change to Courier (my humble recommendation). For visualisation of the text (but not for
> statistics!) Eva is a good candidate since you can use the Eva-hand 1 font to see whether it
> really looks like Voynichese.
> Some questions/observations: -
> as you are probably aware, one existing piece of software that can generate VMs look-alike
> text is of course Monkey itself, but that is a bit like 'cheating', since it needs to read VMs text
> first, before it can do it. Does your program generate the text 'out ofthe blue sky', or can it
> also use a plain text as an input and convert from that? A bit of a leading question, isn't it :-)
> Cheers, Rene
BTW my version of voynichese does start from plain text which is why I am very interested in Brett's program to see how it works and if it can give me a few pointers on the character substitutions. To see if I can make the encipherment have similar letter placings to the VMS.
Brett. I would be glad to email back any details of my own algorithm if you would like to take a look. Do you have access to Delphi? Maybe C++? I could write up the method into a program if it would help.