Hi John --
Yes, Attempto is probably the best generally available parser translator for controlled English so far.
However, per our previous discussions, the grammar-dictionary approach it embodies has severe problems when it ventures outside the research laboratory. In this regard, it's the latest of a long line of superbly researched grammar-dictionary-based systems that have so far failed to make it into widespread practical use. One reason is that the English language is a rapidly moving target -- think of all the new meanings that have arrived with the Net -- the verb "to Google" comes to mind.
It would be nice if the approach did work in practice, since it combines the great intellectual traditions of Chomsky's grammars, and of the Classical logicians, and of John Sowa's fine contributions to the classical logic tradition. However, there's a certain element of "it ought to work so it's going to work" about the approach.
As you know, there's another approach, with far less cultural history behind it, but which happens to work. It's open vocabulary, and largely open syntax, and it's in the system that's online at the site below. It avoids the pitfall of using only classical negation, long ago discounted by Ray Reiter and others as being not the way that databases are actually used. (In an employee database, do you really want to have to list all the people in the world who don't work for a Company?)
I'm sure our debate about the two approaches will continue. But at present, there's one approach which folks think ought to work, and there's a different one which does work. For example, I talked with the Attempto folks about whether they could do the example in [1], and at the time they said that the arithmetic involved was the problem. It's a good challenge problem, but even if someone can do it in Attempto, the grammar and dictionary maintenance ball and chain remains. Plus the fact that authors and users quickly try to step outside any controlled vocabulary and syntax.
Cheers, -- Adrian
[1] www.reengineeringllc.com/Oil_Industry_Supply_Chain_by_Kowalski_and_Walker.pdf
Internet Business Logic
A Wiki and SOA Endpoint for Executable Open Vocabulary English over SQL
Online at www.reengineeringllc.com Shared use is free
Adrian Walker
Reengineering
On Tue, May 6, 2008 at 12:15 PM, John F. Sowa < sowa@xxxxxxxxxxx> wrote:
In various discussions, I have been advocating the use of controlled
natural languages for knowledge acquisition and representation.
One such language is ACE (Attempto Controlled English), which has
been under development for over a decade.
Following is an announcement of the parser and tools (APE) as
open source software. One topic that I discussed with Norbert is
the possibility of generating Common Logic as the output language,
and he is considering it. But anyone who may be interested in
that option should contact him and possibly collaborate on that
effort. In any case, the current output is in Prolog, which is
widely used for commercial AI systems.
Note that this software is being released under the GNU Lesser
General Public License. That option allows the software to be
linked to proprietary commercial software without requiring the
latter to become open source.
John Sowa
-------- Original Message --------
Subject: Attempto Parsing Engine (APE) is now open source
Date: Tue, 6 May 2008 16:37:10 +0200
From: Norbert E.Fuchs <fuchs@xxxxxxxxxx>
To: Tobias Kuhn <t.kuhn@xxxxxx>, Kaarel Kaljurand <kalju@xxxxxxxxxx>
CC: Norbert E. Fuchs <fuchs@xxxxxxxxxx>
Attempto Parsing Engine (APE) is now open source
----------------------------------------------------------------------
Today we are releasing the source files of the Attempto Parsing Engine
(APE) plus some related tools under the GNU Lesser General Public
License. Download the zip-archive from:
http://attempto.ifi.uzh.ch/site/downloads/files/ape-6.0-080506.zip (228 KB)
(Note that all the downloads are also available from the
new Downloads-section on our website.)
These tools are completely written in Prolog, specifically SWI-Prolog
(http://www.swi-prolog.org). There are no other dependencies. In order
to use APE, you must install SWI-Prolog and its packages
(minimally: clib and sgml) first. Recent versions of SWI-Prolog (5.6.46
and higher) should be OK. SWI-Prolog is available in a pre-compiled form
for Windows, Mac OS X and many Linux distributions. Its source code is
also available. See http://www.swi-prolog.org/dl-stable.html. Please let
us know if you encounter problems using APE on your platform.
The distribution includes the following packages:
• parser/ contains the Attempto Parsing Engine (APE) (tokenizer, grammar
files, anaphoric reference resolver)
• lexicon/ contains various lexicon files, notably a content words
lexicon with ~2,000 entries
• utils/ contains various modules, mostly for translating the Discourse
Representation Structure (DRS) generated by APE into other forms
(standard first-order logic syntax, OWL/SWRL, back to ACE, etc.)
• logger/ contains the error logger module
There is an easy-to-use command-line tool `ape.exe' that provides an
interface to APE and the other modules. For example, executing:
./ape.exe -text "Every man is a human." -solo drsxml
will output the DRS of the ACE sentence "Every man is a human." in XML.
To compile this command-line tool, just execute `make_exe.sh' (or
`make_exe.bat', if you are using Windows).
In a separate package we are also releasing a content words lexicon with
~100,000 entries under the GNU General Public License. This lexicon can
replace the smaller content word lexicon that comes with the APE
package. Download the zip-archive from:
http://attempto.ifi.uzh.ch/site/downloads/files/clex-6.0-080506.zip (668 KB)
and just replace `lexicon/clex.pl' by `clex.pl' contained in this archive.
There is more to come. The next release will include more support for
the developers: more documentation of the various modules, our
regression test set with some 3000 ACE-DRS pairs, example code,
instructions on how to set up a web-interface for APE, how to call APE
from Java, etc.
In the future we plan to make even more ACE tools public as open source.
These tools include the ACE reasoner RACE, the AceRules system, the
AceWiki, the ACE View ontology editor, the OWL verbalizer.
Please contact us for support, feature requests, alternative licensing
options, etc. by sending an email to fuchs@xxxxxxxxxx
<mailto:fuchs@xxxxxxxxxx>.
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (01)
|