I think what Rich is getting at is “semantic underspecification”, which arose in the 1980s in natural language semantics, but also in a second wave in the 1990s, initially using mechanisms such as unification (in theories such as Head Drive Phrase Structure Grammar, Lexical-Functional Grammar, various Categorial Grammars), but later type theory and type inference even within the lexicon (e.g., Pustejovsky’s 1995 Generative Lexicon, which uses qualia; Paul Buitilaar’s 1998 PhD thesis, etc.)
Thanks,
Leo
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Rich Cooper
Sent: Saturday, June 11, 2011 10:13 PM
To: '[ontolog-forum] '
Subject: Re: [ontolog-forum] Run, put, and set
John,
I disagree specifically with your belief that run, put, set etc require the representation, storage and interpretation of those 645(?) viewpoints. Instead, I believe that there are many more than those few (645) senses that have been recognized officially by the lexicographers, while people who actually use English all day long probably have far more than that number. But not all at the same time. Only a few interpretations per person are sufficient.
There are only a few thousand words that participate in most daily conversations of Isaid person, even for very learned participant persons. The senses of each of those few thousand words NEED NOT be distinguished if the method of comparing them is to simply order them alphabetically. Then even a simple binary search can find a matching word within log2N time and N space. So the distinguishing among words, even among millions of phrases, is neither difficult nor computationally expensive.
But before alpha ordering sentences, some of the words should be mapped into variables. I use the underscore (_) as the first letter of any variable symbol to make the simple lexical detection of variables easy and immediate, and every new lexical is entered into the local context dictionary. So the phrase
_Actor ran _Modifier the park.
is a sentence which has two variables, and which can be used as a matching mask to compare any phrase containing the word “ran” to all other similarly generalized stored phrases in a signature dictionary. For each match, the Q&Aer has to interpret the new bindings for _Actor and _Modifier, which it will find in a mature system in the same signature dictionary.
Somewhat like syntactic decomposition parsers, but without the large amount of complexity if the signature dictionary has grown to accommodate an acceptable threshold of error reports. To guarantee convergence, be sure that every phrase which has NO MATCH instead has a previously reviewed interpretation function, and be sure that every MATCH has unifiable bindings to existing symbols or to an acceptable threshold number of new symbols.
Because this approach works effectively, even though not as perfectly as mathematically correct parsers could in principle do but presently don’t, it has been surprisingly effective over tens of thousands of documents I have personally analyzed using dictionary constrained contextual decompositions substituted for the more traditional parsing, and instead of converting to a supposed equivalent in FOL. Done in layers, it works also, but that is another topic.
The point is: people use stored representations (i.e. a phrasal dictionary) and layers of pattern recognition (i.e. experiences and specializations thereof) which, accidentally, includes FOL. But FOL is so much less than the diversity of English in that few thousand words per day, that it appears incidental, with logic, math, science and engineering as afterthoughts that showed up after some two hundred thousand years of refinement on the engine that so evolved. We were lucky.
JMHO,
-Rich
Sincerely,
Rich Cooper
EnglishLogicKernel.com
Rich AT EnglishLogicKernel DOT com
9 4 9 \ 5 2 5 - 5 7 1 2
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of sowa@xxxxxxxxxxx
Sent: Saturday, June 11, 2011 6:25 PM
To: [ontolog-forum]
Subject: Re: [ontolog-forum] Run, put, and set
Rich,
I'm not sure what you're disagreeing with.
I agree with your conclusion:
Our software should discuss our needs and then find a way to fill or ameliorate them. To do so requires, IMHO, logic embedded into rhetoric.
The three branches of language are grammar, logic, and rhetoric.
Rhetoric deals with the purpose of language. Without rhetoric, language would be useless.
John