Re: Fw: Intro to natural language processing

["Rich Cooper" <richcooper@mindspring.com>]

Party of Citizens wrote:
<snip/>
> The clinical psychologist takes it scenario-by-scenario. "Given time,
> place and person parameters ____, _____ and ____, what are the examples of
> 'appropriate' or normative natural language?" Imagine we have most of the
> time, place and person scenarios listed for people in our culture; and we
> have natural language norms for those scenarios. Clinical psychologists
> and all of those related professionals act as if they have such scientific
> information. The entire DSM taxonomy used by clinicians the world over
> would fall apart if those norms were invalid. What would we find in the
> tables of norms if they were on hand? Strings of letters and words.
>
> As long as I emit letter and word strings which fit into those tables of
> norms, what can a clinician say except that my natural language is normal?
> Does it matter if I am a naughty robot like Bender of the Futurama
> television series and have no "comprehension" of what I am saying in the
> metaphysical sense? No. "Bite my shiny metal ass" seems to be normative
> for a variety of time, place and person scenarios in the year 3,000 AD.

Encoding the DSM into machine processable categories is not too hard,
and I vaguely remember a project to do so back in the 80's or 90's.  But
the problem seems to be mechanizing the interpretation of DSM entries.

Psychologists have a great deal of jargon, just like any other domain,
and that jargon is very ambiguous, at least to algorithms designed to
select 'appropriate' behaviors from 'innappropriate' ones.

But generally speaking, I think you have the right idea about representing
scenarios within which the encoding can begin.

JMHO,
Rich


<snip/>