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

Re: LSC and the VMS

On Sat, 22 Jan 2000, Mark Perakh wrote:
> Moreover, if VMS symbols are not individual letters, all LSC results hang
> in the air. Best, Mark

In order to draw meaningful conclusions from statistics, we really need
statistics that are stable when we change the arbitrary decisions (such as
where letters begin and end) - as I pointed out on this list in Nov 1998.
Unfortunately, I haven't been able to come up with any such statistics,
but I do have some hopes for LSC.

The fundamental problem is that we need to compare large-scale
correlations with small-scale correlations, *without* knowing exactly how
large or small the scales are except in relation to each other.  I mean,
if I take one "word" of the VMS, I don't know if that's really the size of
linguistic unit we normally call a word, if it's more like a "letter", or
if it's (like the Masonic reminder books someone mentioned, they have a
proper name which I forget) it's an acronym that could abbreviate a whole
sentence.  All I know is that there are one hundred VMS "words" in a
passage one hundred VMS "words" long.

To be useful, statistics have to look at relative scale instead of
absolute scale, and the LSC looks good to me because it seems to do that.

Matthew Skala                       "Ha!" said God, "I've got Jon Postel!"
mskala@xxxxxxxxxxxxxxxxx            "Yes," said the Devil, "but *I've* got
http://www.islandnet.com/~mskala/    all the sysadmins!"