Re: VMs: VMS generator table

[Gabriel Landini <G.Landini@xxxxxxxxxx>]

On Friday 05 December 2003 01:08, Jeff wrote:
> Find it here http://www.georgeboolefanclub.com/vmstextgenerator.html

Sorry if this sounds disappointing, but that is an unnecessarily complicated 
and not accurate transition probability table (given 2 characters, generate 
the third one).

Why "unnecessarily complicated"? Because there is a better way to encode SMX 
by using the space character. The space tells you whether you are at the 
start or end of a word.
Why "not accurate"? because the same weight is given to all the transitions.

If you followed the suggestions posted so many times  to read the mail 
archives AND Bennett's book (which I suggested to you not too long ago) you 
would have saved yourself a lot of time.

Jacques' MONKEY can do this type of text generation in a more accurate way.
If done properly (as with MONKEY) the approximation to the VMS is only valid 
for the order n length, but not beyond (i.e. if one is generating the third 
character the result shows the right proportion of triplets (they are all 
valid ones) but not 4-plets, 5-plets etc.).
That is why it is so easy to spot that your generator is not accurate. Any 
word with more than 3 characters may have a "never seen" combination of 
characters.
You could try order making higher order tables, but I still suggest to read 
Bennett's book before spending any more efforts on this.

Cheers,

Gabriel


______________________________________________________________________
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list