Re: Database duplicate
Linas Vepstas
>
> Is there any documentation explaining how OR works then?
> http://www.link.cs.cmu.edu/link/dict/introduction.html implies that OR is an
> exclusive OR.
Well, its exclusive, but its not a boolean conjunctive!
Any of the first few papers on link-grammar provide a much better
theoretical introduction than the website. Also, the cmu website
is obsolete, there are significant changes to the documentation,
and these can be found only on the abiword website.
I'mm cc'ing the mailing list, as all this is presumably of general
interest.
> Another place that doesn't make sense caught my eye:
> <directive-opener>:
> {[[Wi-]]} &
> ((Xc+ & Ic+) or
> ({Xd-} & (Xc+ or [[()]]) & [[COa+]]));
>
>
> no.ij nope.ij nah.ij no_way yes.ij yeah.ij yep.ij yup.ij
> ok.ij okay.ij OK.ij fine.ij exactly.ij sure.ij whatever.ij
> hah.ij hey.ij well.ij:
> <directive-opener> or
> Wi-;
>
> Wi- exists in both sides of the top level OR. So the first part of the top
> level OR can only work when there is no Wi- link, but it also contains Wi-
> link clause. What does this mean? Shouldn't just one of them be dropped in
> case of no.ij and similar words?
Well, <directive-opener> or Wi-; says that hah,ij can connect using Wi-
and only Wi- (and nothing else). Alternately, it can connect using
Xc+ & Ic+. If it needs to, then it can connect using Xc+ & Ic+ & Wi- but
this second form is strongly discouraged (the double [[]] brackets) which
means that it will be used only if something less costly isn't found.
In other words,
Wi- or {[[Wi-]]} & ((Xc+ & Ic+) expands into
Wi- or (Xc+ & Ic+) or [[Wi- & Xc+ & Ic+]]
the exclusive-or nature of or means that only one of these
can be applied at a time.
--linas