The new draft is now complete with abstract and references:
http://code.google.com/p/genifer/downloads/detail?name=logica-universalis%20%2813-July-2012-3%29.pdf&can=2&q=#makechanges

On Mon, Jul 2, 2012 at 3:34 AM, Matt Mahoney <[email protected]>wrote:

I'm not sure what the point is. You introduce a mathematical formalism
> for describing parse trees. What does this buy you? What problems can
> you solve now that you couldn't solve before? What is the motivation?
>

It does not make inference *easier*, but it provides a formalism that is
universal for natural languages.  The status quo is that there is no
logical representation that is universal for NL, so we cannot even begin to
perform inference for general NL texts.  My new logic hopefully makes this
possible.

I assume it is to do inference. Maybe you could give an example where
> the problem is easy using your notation, but hard given a normal parse
> tree. For example, a parse of "big red house" would maintain the word
> order, where in your notation, the words "big" and "red" commute.
>

It doesn't make a *particular* inference easier, but it is general enough
for any inference involving NL.  Also it does not handle the kind of
nuances about word orders (yet).


> The problems of parsing natural language are well known, such as
> ambiguity and grammar complexity. I don't think you claim that your
> formalism makes any progress here. For example, "red big house" is
> grammatically correct and semantically equivalent, but still unusual.
> I don't see that it makes the problem of inductive learning any
> easier.
>

Once we have a universal representation of NL, we can start writing
learning algorithms.  In the part this task was hampered because the
logical forms for NL were even more awkward to use.

I guess finishing the abstract would help. I know abstracts are hard
> to write, but I take that as a warning that I need to re-think the
> rest of the paper. It is easy to say "a new formalism is introduced",
> but harder to say why.
>

Thanks for the suggestion, maybe I'll add a few words about that in the
paper (not submitted yet)....

KY



-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com

Reply via email to