Re: nature -> "human brain" -> "language terms" ==>> knowledge ?
On Sunday 20 March 2005 00:57, Rich Cooper wrote:
> Rob Freeman wrote:
> <snip/>
>
> > I am proposing that language can be a representation for an infinity of
> > meaning, because you can find an infinite number of patterns among
> > utterances
> > in a language.
> >
> > But I think those patterns _must_ be found "on the fly" because they will
> > be
> > inconsistent with each other (which is indeed what we observe.)
>
> I like this approach. Search engines look for words and phrases and create
> economic value from that approach. Chat engines, weak as they are, provide
> some entertaining value, though they certainly aren't strong enough to
> model what people do.
>
> So wading through a lot of text and finding only those parts that make some
> sense seems to work better than any of the machine translation or Q&A
> systems that try to force a formal representation around all the
> utterances.
We can do better than that, Rich. We don't need to limit ourselves to finding
parts of texts which make the best sense. We can rearrange texts to make new
sense.
Once we understand the meaning is in the way we order examples, it is just a
matter of figuring out meaningful ways to interpret that order. Whether in
action, or description.
I have worked with finding the order indexed by a given sentence, essentially
parsing it. I don't know how you would go about finding a sentence given the
order, the sort of thing you might need to do to describe an image, or
provide a precis of a news article, but I am sure it can be done. Something
like starting with the results of a Web search and inducing what the likely
search query was.
True Q&A surely requires some kind of agent working on top of that. Something
to interpret your request and decide on the best response. To do that you
need more than just a model of meaning, you need something which thinks. But
it is sure that a flexible model of meaning is the first step.
-Rob