[Attempto] Broader the range of texts that can be mapped to ACE

[Simon Spero]

On Sun, Jan 27, 2013 at 12:15 PM, Norbert E. Fuchs <fuchs at ifi.uzh.ch> wrote:


> In my view, natural English and controlled English are conceptually
> different languages that need a human translator.
>

I am not sure if this statement is too weak or too strong; that's the
problem with natural languages :-)  A lot depends on what one means by
language, different languages, and conceptually different languages.

I would claim that any controlled version C of a natural language L must be
a subset of L,  such that every sentence in C is a sentence of L;  that
every sentence in C has a single parse, and that this parse must be a
possible parsing of the sentence in L.  I would probably extend parse and
sentence to DRS and text (anaphora, pronominal co-reference , etc) but I'm
not sure if this a necessary condition.

A designer of a controlled version of a natural language is constrained by
the grammar of the natural language; they  are not free to introduce new
grammatical constructs, etc., should recognize/generate as natural a set of
sentences as possible, but must avoid ambiguity.   This act of controlled
order (taxis)  is conceptually different from the grown order of natural
languages (kosmos), but I'm not sure if it is the languages themselves are
conceptually different.   Certainly the tasks of designing a Controlled
language and designing a Constructed language  are unrelated.

It could be argued that  by some definitions, a Controlled English (e.g.
ACE) is a dialect of Uncontrolled English; texts in ACE are understandable
by speakers of English; however, these speakers need some training or
guidance to produce correct ACE.

It's also not clear for which cases  a human translator is needed; I will
use the example below to illustrate some places where existing automated
system can solve the problem.


> For a comparison take a text problem: *A farmer has a number of cows and
> a number of ducks. Altogether the farmer has 100 animals that have 260
> feet. How many cows and ducks does the farmer have?*
>
> Notice first that you need to add knowledge, for instance that cows and
> ducks are animals, that no cow is a duck, and that cows have 4 feet and
> ducks 2 feet.
>


This knowledge is available to CyC, and is retrievable extremely
efficiently.  There is other meta-knowledge that it also has; e.g. that it
doesn't know of any animals each of which has 260 feet (although in
different  circumstances it might be forced to hypothesize mutilated
millipedes, or mutant ducks).

If the genre of the discourse is identified as being a mathematical word
problem, then other necessary pragmatic knowledge can be brought to bear;
that the only salient kinds of animals are those mentioned (cows and
ducks); that every cow and every duck has the default number of legs*, that
the problem is solvable, and solvable using mathematical techniques, etc.

Notice second that in order to arrive at the mathematical expression of the
> problem – namely the linear equations X+Y=100, 4*X+2*Y=260, where X is the
> number of cows and Y the number of ducks – you need to do some abstraction
> and need to find a strategy how to solve the problem.
>

If the genre is identified, then this part is not to hard;  STUDENT did a
fairly good job of this part, even most of the natural language
understanding parts of the project were not really generalizable.   I don't
know if anybody ever added a STUDENT like reasoning module to CyC, but it
would not be that hard.

That doesn't take away from the fact that most of the knowledge in CyC had
to be encoded by humans. This knowledge had to be expressed in an
unambiguous  logical form.

Simon

* A cow like that, you don't eat all at
once<http://www.jokesplace.com/joke/famouspig.html>.
 See also: Cattle Mutilation, and Abduction
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ifi.uzh.ch/pipermail/attempto/attachments/20130127/c2a3d387/attachment.html>