[Top] [All Lists]

Re: [ontolog-forum] Semantic Systems

To: "'[ontolog-forum] '" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "Rich Cooper" <rich@xxxxxxxxxxxxxxxxxxxxxx>
Date: Mon, 29 Jun 2009 09:34:29 -0700
Message-id: <20090629163520.264A1138ECA@xxxxxxxxxxxxxxxxx>

Hi John,    (01)

<snip/>    (02)

JS> Pat Hayes was talking about the universe of all possible statements
that one might consider about a given set S of individuals.    (03)

JS> As an analogy, consider a relational database, which consists of a
collection of tables, which can be described using only two operators:
existence and conjunction.  That DB could be considered a collection
of observation statements about Pat's set S of individuals.    (04)

JS> Pat noted that the totality of all true FOL statements about those
individuals is countable.  Each of those statements can be shown
to be true by evaluating their truth values in terms of the tables
of the DB -- i.e., you can ask arbitrarily many SQL statements
about that DB which evaluate to true by an SQL query system.    (05)

JS> I don't understand how that relates to your question.    (06)

JS> John    (07)

RC> I'm trying to organize sample sentences in a database in a way that
meets the following requirements:    (08)

Req 1.  Canonical semantic template representations to be stored and
retrieved from the DB where the variables are not part of the template, but
the signature of the template is canonical.  Therefore the templates must be
retrievable due to semantic structure.      (09)

Req 2.  The templates must match much of common English usage for a given
application domain.  Given a set of training sentences, I want to record
them in a way that can be matched against new input sentences.  That is,
they provide an interpreter with templated meanings.  The sentences are
conversational, such as a person might have with a simple friend.      (010)

Req 3.  The LGP and a version of dependency grammar for English will be used
to translate input sentences into phrase templates with variable slots for
subjects, verbs, objects, adjectives, adverbs and other LGP concepts.      (011)

Req. 4. After matching an input sentence against the various templates, the
software must reconstruct the meaning of the speaker and simulate performing
relevant actions in a simulated world.      (012)

I keep going back to Terry Winograd's QA system with blocks world
simulations.  His dissertation describes how he had essentially no tools to
help organize his conceptual structures.  If there were a better way to
build such programs, they might bridge a gap in the NLP technologies of
today.  My talking things (GPS, computer, phone, ...) are so incompetent at
English, that I would like to raise that bar just a little.  I want to
produce a QA system structure that can be applied to various domains very
easily.  To do that will require an extremely complicated, but well
engineered, set of algorithms and representations.      (013)

So that's my purpose here.  How do I make a easily reusable factored English
QA system shell?    (014)

-Rich    (015)

Rich Cooper
Rich AT EnglishLogicKernel DOT com    (016)

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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (017)

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