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

Re: VMs: Number crunching the Fincher window

Zitat von Rene Zandbergen <r_zandbergen@xxxxxxxxx>:

> Hello Elmar,
> a different type of statistic... nice.
> But did you use the Eva transcription?

Yes, it was done in EVA.

> Assuming that: in, iin, iiin are single characters
> and being quite sure that ch and sh are as
> well, the stats better be based on a Currier- or
> FSG-like transcription scheme. 
> It should change the statistics in the shorter-
> length area...

I'd thought it wouldn't really matter if I wanted to test the Fincher 
hypothesis. (I guess it would be reasonable to assume that, unless the VM 
author used a square grid paper, even his window captured a few characters more 
or less each time.)

My initial plan was to find the "master sequences", and thereby reengineer the 
whole thing. My little script captures not only the various sequences, but also 
the frequencies of the letters following them. So what you can do is produce a 
string of the "most probable" sequence, which should dedicatedly follow a 
master sentence, until you hit the end of that sentence. At this point, 
frequencies for the following letter should be more evenly distributed, since 
the window has to pick up again on a random spot on the chart.

Unfortunately, at this point ugly facts begun to rear their heads, because, 
looking at some 20000 sequences, I'm a bit stumped on how to assess them with 
either reasonable manual effort, or limited programming time.

My data pool consists of all existing sequences (lengths 4 through 16 EVA 
chars), the number of occurences for each sequence, and lists for the 
frequencies of the letters immediately following the sequences for each case.

Any good ideas?


(Yes, I'm quite excited that, for the first time, we really seem to be having a 
theory we can put to the test.)

debitel.net Webmail
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list