Re: Unsupervised learning of natural languages
From: | Gary Shannon <fiziwig@...> |
Date: | Thursday, November 3, 2005, 20:24 |
--- tomhchappell <tomhchappell@...> wrote:
<snip>
>
> Also, the Context Window can be set wider. In their
> test runs, it
> was always 5 or less -- I guess that would be as if
> all their rewrite
> rules would have had, at maximum, the form
> ABCDE --> FGHIJ
> where each of those letters could be any arbitrary
> grammar symbol (or absent) -- a terminal or a
> non-terminal -- and not
> necessarily distinct from each other.
Bear in mind that after re-wiring the paths the length
of the string of nodes becomes smaller. For example,
ABCDEFGHIJKL where (DEFG) becomes Q -> ABCQHIJKL, and
then (CQHI) becomes R so that the string is then
ABRJKL, now the context widow at width 5 encompasses
they tokens A and K which were originally 10 step
apart. In this way, as the structures are recursively
collapsed, distant relationships become nearby
relationships and are eventually snagged by the
context window. Or at least they CAN potentially
becomes nearby relationships.
--gary