[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: CE Starter grammar
. Further comments interspersed below, prefaced "GH2>>>
Cheers Graham Horn
National Data Standards Unit
Australian Institute of Health and Welfare
Email: Graham.Horn@aihw.gov.au <mailto:email@example.com>
From: John.Velman@HSC.com [mailto:John.Velman@HSC.com]
Sent: Tuesday, June 05, 2001 7:02 AM
To: Horn, Graham
Subject: RE: CE Starter grammar
Thanks for the clarification and comments. Some response interspersed
-------John V wrote (comments by GH embedded -- hope you can sort it out):
>Well, both Matthew and Graham are right. We need to focus on something
that looks like English hence the need to explicitly establish a syntactic
structure consistent with traditional structures. Our goal is a language
stands unambiguously in an isomorphic relation to KIF.
>Perhaps one approach is to take a minimal syntactic structure such as Adam
>GH> I developed a bigger structure than necessary for our immediate
purposes in order to ensure sound extensibility.
>GH> If you look at the structure I put forward, you'll see it overcomes
some shortcomings I found in Matthew's original. Furthermore, I made the
------------------------------^^^^^^^ (Best viewed with fixed pitch font!)
JRV: Did you mean Adam's original? If Matthew has suggested a syntax for
SUO-CE I've missed it.
GH2>>> Yes, Adam's. Sorry, I was probably reading a reply.
>changes because There were some pretty basic things I couldn't represent
>GH> The main point I suggest is that we separate the word types from the
function types where those words are used. For example I added the concepts
of process, qualifier, and adverb. I also couldn't see why direct and
indirect objects should be treated as part of the verb. I further addressed
some optionality considerations.
JRV: The distinction between grammatical function and logical function is a
important one, and perhaps difficult to deal with. I don't think Adam
suggested treating an indirect object as "part of the verb." But it is
rather standard, as far as I can tell, to call the top level parts of a
declarative sentence "noun phrase" and "verb phrase" I.e.,
JRV: The verb phrase is further decomposed including noun phrases (e.g.,
indirect objects phrases -- also called verb complements) as constituents.
I'm, by no means an expert on this stuff, and if I've missinterpreted your
comment, please set me straight.
GH2>>> You've interpreted me correctly. At this stage, I can't see what is
gained by rolling the direct & indirect objects up with the verb. Did you
get an intelligible version of the hierarchy I was proposing? I see
considerable advantage with keeping the "noun" parts, with their adjectives,
etc., separate from the "verb" parts with their adverbs, etc.
>agree on some core terms and syntax (in CE) for "exists" "forall" "and"
"or"... and then play with it a little. My experience suggests that it is
really useful to play with some simple -- even toy -- examples before
getting too deep into a design. (If I have some time later on today - and
noone else has done it -- I'll give this a try.).
JRV: I did spend a little bit of time (unfortunately, I didn't have much
time to spend on this over the weekend either) trying to write down some
alternative English translations of the KIF in Chris Menzel's document
giving KIF syntax and structural ontology.
JRV: This has been very interesting, and has shown me what some of the
problems may be, but I don't have any solutions or even questions to share
GH2>>> When you get to it, I will be interested in seeing the problems you
>Also, while I agree that it will ultimately be useful to expand into
questions and commands as Graham suggests, I propose that for now we
concentrate on assertive statements.
>GH> I think you're reading too much into what I was doing. I was just
trying to remove some problems I could see with the structure presented. I
admit, I responded by presenting a far more comprehensive structure, but I
needed to go to that level in order to satisfy myself that what I was
proposing had the soundness I believe we need for this sort of project. I
was also drawing on a pretty well established paradigm.
JRV: Yes, I was probably reading more into it than you intended. I agree
that it is useful to present a larger context than the area actively under
JRV: I also dug out the ATEMPO Controlled English documents and papers. We
ought to try to benefit from what is available there. I think that what we
need to accomplish is simpler, because I don't see the SUO as being a
'discourse' in the same sense as the specifications that ATEMPO CE is
intended to deal with.
GH2>>> I don't know enough of ACE to be able to comment. I agree we should
just concentrate on what we require (so long as we have a sound structure
JRV: For those who are interested: Covington's book on Natural Language
for Prolog Programmers contains a discussion of an Earley parser (and the
prolog code is available on the book's website). If you haven't already
tinkered with Prolog, you probably won't want to tackle this, however. [
remainder of old messages snipped]
GH2>>> I haven't already tinkered with Prolog.