SUO: RE: Model of Activity and Action in SUO ontology
John F. Sowa wrote:
> Rich,
>
> I don't know of any work that has been done to link
> Anna W's primitives to other work on ontology.
>
> But in various writings, she has expressed misgivings
> about using logic to represent the semantics of natural
> languages. That doesn't mean that other people can't
> use logic to represent her categories, but it indicates
> that she would be unlikely to sympathize and collaborate
> with such efforts.
Yes, I've emailed both Wierzbicka and Goddard twice with
no responses yet.
> Re Anna W's primes: They are a very interesting attempt
> to represent a wide range of concepts, but she blurs many
> important issues in her classification. I wouldn't claim
> that her so-called "primes" are truly primitive, and they
> must be further analyzed before they can be related to
> other work in logic and ontology:
>
> 1. Indexicals: She uses the word "this" as a general
> indexical. Other primes must also be distinguished
> as indexicals with qualifications: I, you, here,
> there, now. In CGs, I represent them with the marker
> # and a type label and/or other qualifier. For
> example, "I" could be represented as [Speaker: #]
> for "the speaker" in a discourse, but there is a
> lot more to be said in general.
>
> 2. Quantifiers: someone, something, and all represent
> the logical quantifiers with additional qualifiers.
> Other terms, such as one, two, many, and much, are
> generalized quantifiers. There is a lot more to
> be said about these.
>
> 3. Interrogatives: when, where, who, what, why, and
> how are the basic interrogative words. In CGs, I
> represent them as [Time: ?], [Place: ?], [Person: ?],
> [Entity: ?], [Reason: ?], and [Manner: ?].
>
> 4. Modals: The word "can" is a modal marker for
> possibility, which may be combined with qualifiers
> to represent various modalities. Verbs like
> "want" and "think" also have modal or intensional
> effects, which must also be recongized.
>
> In short, I would not adopt Anna W's primitives exactly
> as she expressed them, but I think that she has done
> a lot of hard work in analyzing other concepts in terms
> of them. I would start by translating her primitives
> to CGs and then using that translation to map her other
> definitions to conceptual graphs. Then the CGs could be
> compared with the definitions used in other ontologies.
>
> John
I agree. But I'm using her primes in characterizing
phrases with human semantic concepts. As a sentence
gets parsed, each phrase needs some interpretation, and
her primes are the best available choice I've found so
far. Much of the linguistic research is grammar-oriented
(subject, noun phrase, complement, verb phrase, ...) with
no clear semantic correspondence.
In semantics, people use very mathematically oriented
predicates that don't sound intutitive to me. Since people
were using language for at least the last 200,000 years,
and since mathematics is only a few thousand years old,
I'm opting to use the premathematical notation that
Wierzbicka came up with instead.
That doesn't mean I won't also add more mathematically
consistent predicates later, but I think there needs to
be another level between syntax and semantics as presently
practiced. There's just too big a gap from what we say
and how we think and feel about what we say.
For precision, there is still a set of domains in my work.
They include the usual suspects (Integer, Extended, Boolean,
String, Functor, object, ...) but I want to add a phrase type domain to
the list. By using Wierzbicka's 63 primes, I can classify
phrases along that intuitive axis, and perhaps that will
be useful later in actively interpreting the phrases,
a la Winograd's blocks world.
I guess my basic motivation is that Winograd's simple
example shows that there must be both procedural and
declarative forms of a sentence if "understanding" is
to be achieved. Maybe Wierzbicka's 63 primes will add
a reasonable link between sentence and action. Maybe it
will take a great many more primes than she used in
her own work. But at least it gives me a reasonable
starting point to work from.
Hopefully it won't have to be abandoned down the line,
but there is always that risk in breaking new ground.
Rich