[Top] [All Lists]

[ontolog-forum] Controlled Natural Language [was Apology...]

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Tue, 04 Nov 2014 21:26:50 -0500
Message-id: <54598AEA.8000409@xxxxxxxxxxx>
David, Joseph, and Rich,    (01)

I'll group the responses under the CNL heading, since it is the
most relevant to Ontolog.  Siri (and similar systems) support a
very flexible combination:  speech recognition + search engine
+ CNL for output (help facility) + ELIZA-like pattern recognition
for responding to greetings and casual remarks.    (02)

Systems like Cyc can be added to support more complex reasoning
for various purposes.  Each addition would add a new "module"
to support new functions, but the old functions would still be
available.  No one would have to learn a *new* language.  It
would always be controlled English (or other NL) -- but with
gradual improvements in what the system can understand and do.    (03)

In a previous note, I mentioned OACES, which provides a highly
structured English-like dialect (ACE) that can be mapped to and
from first-order logic.  With speech recognition and generation
added to ACE, it could make a Siri-like system into a tool for
answering questions about the ontology (i.e., definitions). It
could also support semi-automated ways for adding new definitions
or revising old definitions.    (04)

> Absolutely no argument that Cyc's goal of supporting ordinary
> language is a worthy, laudable & valuable goal.
> And I would continue to argue there is a HUGE amount of language
> that is far from "ordinary" & far, far removed from any sort of
> formal dictionaries.    (05)

I agree with both of those points, but with some qualifications:    (06)

  1. No system today can "understand" unrestricted NLs.  But good
     guessing, speech recognition, and a simple way to correct
     misunderstandings can support a usable interface.  (See below
     for excerpts from the previous Siri dialog.)    (07)

  2. By 'ordinary', I meant the usual highly flexible way of moving
     from one topic to another by subtle shifts in the meanings of
     words.  This is the antithesis of attempts to specify a fixed
     meaning or set of meanings.    (08)

> but what about jargon, slang, acronyms, salacious, industry
> specific, work group specific, etc., etc., etc?    (09)

There's a continuum.  A pattern-matcher can match any kind of
patterns.  A parser can handle any vocabulary you put in its
dictionary.   When you speak ACE, it recognizes ACE.  If the
ACE parser fails, the system switches to a more flexible
pattern matcher.    (010)

You can add folksonomies and special modules for any purpose.
 From the user's point of view, the system is always using the
same NL, but its vocabulary can adapt to different tasks.    (011)

> Any pointers to introductory tutorials?    (012)

See ACE resources: http://attempto.ifi.uzh.ch/site/resources/    (013)

But remember that these are research projects developed by
graduate students.  Some of them can be used as building blocks
for creating other combinations.  One example is APE, the ACE
Parsing Engine for translating ACE sentences to DRS.  Another
is ACE View, a graphic plug in to Protege.    (014)

> It would be good to find out how [Norbert Fuchs] put the DRS
> [for ACE] into his FOL output automatically.  Do you have any
> links  to the DRS work he did?    (015)

DRS is a version of FOL that is isomorphic to Peirce's existential
graphs.  You can use APE to translate ACE to DRS.    (016)

In the following paper, I show how DRS maps to EGs and
then to conceptual graphs and predicate calculus:    (017)

    http://www.jfsowa.com/pubs/eg2cg.pdf    (018)

For DRS examples and their translations, see Figures 9, 15, 16,
and 18.  Figure 17 applies Peirce's rules of inference to the EGs
in Figures 15 and 16.  Exactly the same rules in exactly the same
sequence can be applied to the DRS to derive the equivalent result.    (019)

And note:  I have been talking about ACE because you can download
the basic tools for free.  But there are many other versions of
CNLs with varying advantages and disadvantages.    (020)

John    (021)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J    (022)

<Prev in Thread] Current Thread [Next in Thread>