John B and Kingsley, (01)
Before commenting, I'd like to mention CogniToy, a talking dinosaur
that connects to IBM's Watson. To start, parents connect to the
CogniToy web site, type some data about the home wi-fi, the child's
name, age, etc.: http://www.wired.com/2015/02/cognitoys-ibm-watson/ (02)
This illustrates controlled NLs for output (by the toy), a very
forgiving NL for input, and learning methods for automatically
adjusting the dialog to the human. With a different vocabulary,
such an interface could be adapted to people and tasks of any kind. (03)
JB
> you may be throwing out the baby with the bathwater. When you
> throw out the "smelly words" you are also discarding their data
> structures. That leaves a void that needs to be filled. (04)
I also said that legacy systems are *successful* systems. I
would never throw out anything that has proved to be useful. (05)
KI
> Our challenge remains getting narrative control back from buzzwords (06)
Yes. I was not complaining about the useful systems, but about
the metalevel verbiage used to describe them. Since every digital
computer is a logic machine, every term that describes what it
does can be defined in logic. (07)
JB
> the issue is not with the word itself, it is with the lexical
> load that the word brings with it in its use. This cannot be
> remedied with logic, it is a CS issue with how data structures
> are used, labeled and linked. (08)
I completely agree with the first sentence and with the final
clause. The only point of disagreement is the claim that
the problem cannot be remedied with logic. (09)
I realize that a huge amount -- in fact the great majority --
of what people say and do cannot be specified in logic. But
when we talk about what a computer does, the words we use can
and should be precisely defined. (But I wouldn't include the
the expletives that people utter when they see what happens.) (010)
JB
> My preference is for a definition and usage of terms that move
> us from <entity> to something closer to a Natural Language. (011)
I agree with that goal. I also believe that steps A to D in your
note could be adapted to systems with an interface like CogniToy.
But there are two *disjoint* kinds of vocabularies: (012)
1. The content words of the many nomenclatures, terminologies,
vocabularies, folksonomies, hashtags, etc. In general,
these words would include anything that any human might use
and any Siri-like bot might reply with. (013)
2. The metalevel terms that we (Ontolog Forum and other
developers) use to describe, design, and implement
the systems that support those interactions. (014)
The terms in #2 that describe how the systems are designed must be
defined with sufficient precision that they can specify what the
computer system does. That implies logic (formal or informal). (015)
The word 'entity' does not belong in #1. But the word 'thing' is
confused and confusing when used in #2 for an ontology. (016)
My first choice for the top of an ontology is a sans-serif T with
the associated phrase 'universal type' or just 'top'. For the
bottom, I suggest an upside-down T with the phrase 'absurd type'
or 'bottom'. Those symbols are widely used in the DL community. (017)
Both of those symbols (and the phrases) can be defined in logic:
the universal type is true of everything, and the absurd type
is false of everything. The word 'thing', as used in ordinary
English, has no precise definition -- a large dictionary lists
a few dozen word senses. (018)
KI
> Tables to RDF Property/Predicate graphs aren't about forced
> conversions, implicitly. (019)
Yes, many DB vendors support both SQL and SPARQL access to
the same data in real time. That supports interoperability,
which may continue indefinitely for mission-critical data. (020)
But some people have promoted the one-way mapping of RDB2RDF,
and some unhappy users were subjected to a forced conversion. (021)
John (022)
_________________________________________________________________
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 (023)
|