D7 - Which languages are better than OWL?
Dear all,
I made many changes to the semi-formal summary related to
this discussion (http://www.phmartin.info/suo/ or
http://krl.wiki-site.com/index.php/Main_Page).
First, I added the following statements and an argumentation
hierarchy for them:
- "RDF+OWL is not a good general_KRL";
- "Common Logics is not a good general_KRL" (that one is
deliberately provocative to spur some contributions; I
only regret that CL is not as expressive as KIF).
I include that argumentation hierarchy below.
Please give objections to the statements you disagree with,
or complementary arguments where mine are insufficient.
Second, I organized most of the statements into an
argumentation/specialization hierarchy with the following
statement at the top:
"RDF+OWL is not a good KRL_for_exchanging_knowledge".
This is in-line with the purpose of this discussion.
Third, since the recent exchanges relating to n-ary
situations/relations showed that the initial formulations
were inadequate, I restated them and added objections.
Here is an extract (see below for more details):
"in 2007, no W3C_language is a good_general_KRL"
argument:
- "in 2007, no W3C_language provides a way to specify in
a concept definition that some relations are mandatory"
objection: "OWL permits to specify that a relation is
mandatory via the use of cardinality restrictions
or by specifying it as functional"(pm)
Gian, please check that the related statements below express what
you meant.
Another discussion will begin if there is no further additions
to this one.
Philippe
====================================================================================================
2.2. The more expressive the KRL, the better?
Warning: the sentences in this sub-section are not yet normalized.
"RDF+OWL is not a good KRL_for_exchanging_knowledge"
argument: "RDF+OWL is not a good general_KRL" //details below
(pm, objection: "exchanging_ontologies_via_OWL does not require that
`OWL is a good general_KRL'"(jrSchoening) //details below
);
"exchanging_ontologies_via_OWL does not require that `OWL is a good general_KRL'"
argument: ("exchanging_ontologies_via_OWL is not detrimental to
knowledge_representation_or_inferencing_with_a_language_more_expressive_than_OWL"
objection: - "exchanging_ontologies_via_OWL may lead to a loss of semantics"(pm)
- ("some applications require (or provide better results with)
precisions that cannnot be represented via OWL"
generalisation: "applications requiring precision need
the expressivity"(pm)
)(pm)
)(jrSchoening);
"RDF+OWL is not a good general_KRL"
argument: ("RDF+OWL is not expressive enough for general_KR"
argument: "applications requiring precision need the expressivity"(pm),
argument: - "RDF+OWL does not include F.O.L., general meta-statements nor sets"(pm)
- "F.O.L., general meta-statements and sets are necessary to represent
certain natural language sentences or certain other information"(pm),
argument: ("a general_knowledge_interchange_format should be as expressive
as possible"
argument: "the more expressive the KRL, the better for
knowledge re-use"(pm) //details near end of section
)(pm),
argument: "most practical_applications do not require a language as
restricted as OWL"(gpZarri) //details at end of section
argument: - "the model of RDF+OWL does not permit to define
numerical quantifiers"(pm)
- "a good general_KRL should permit to define numerical quantifiers"(pm)
argument: - "the model of RDF+OWL does not include numerical quantifiers"(pm)
- "a good general_KRL should include numerical quantifiers"(pm)
)(pm), //details below
specialization: "in 2007, no W3C_language is a good_general_KRL"(pm); //details below
"in 2007, no W3C_language is a good_general_KRL"
argument:
- ("in 2007, no W3C_language provides a way to specify in a concept definition that
some relations are mandatory"
objection: "OWL permits to specify that a relation is mandatory via the use of
cardinality restrictionsies or by specifying it as functional"(pm)
)(gpZarri (pm)) //according to gpZarri as interpreted by pm
- "a good_general_KRL should provide a way to specify in a concept definition that
some relations are mandatory"(gpZarri),
argument:
- ("in 2007, no W3C_language provides a way to specify that some relations are
contemporaneous"
objection: - "RDF permits to use meta-statements"(pm)
- "meta-statements permit to specify that some relations are
contemporaneous"(pm)
)(gpZarri (pm))
- ("a good_general_KRL should provide a way to specify that some relations are
contemporaneous"
specialization: "a good_general_KRL should permit to use meta-statements"(pm)
)(gpZarri);
"Common Logics is not a good general_KRL"
argument: - "Common Logics does not permit to define numerical quantifiers"(pm)
- "a good general_KRL should permit to define numerical quantifiers"(pm),
argument: - "Common Logics does not include numerical quantifiers"(pm)
- "a good general_KRL should include numerical quantifiers"(pm);
//note: KIF does not include numerical quantifiers but permit to define them
"a good general_KRL should permit to define numerical quantifiers"
argument: ("a user should be able to represent numerical quantifiers"
argument: - "numerical quantifiers are needed to represent
natural language sentences or certain other information"(pm)
- "applications requiring precision need the expressivity"(pm)
)(pm);
"a good general_KRL should include numerical quantifiers"
generalization: "a good general_KRL should include as many knowledge representation
constructs (such as numerical quantifiers) as possible"(pm);
"a good general_KRL should include as many KR constructs (such as numerical quantifiers)
as possible"
argument: ("a user should be able to use KR constructs (such as numerical quantifiers)
without having to define them"(pm);
"a user should be able to use KR constructs (such as numerical quantifiers)
without having to define them"
argument: - "very few inference engines are able to exploit the logic definition of
complex KR constructs (such as numerical quantifiers) for
inferencing purposes"(pm)
- "it is complex to implement an exploitation of the logic definition of
complex KR constructs (such as numerical quantifiers) for
inferencing purposes"(pm)
- "it is easy to calculate a generalization relation between expressions
using complex KR constructs (e.g., quantification expressions such as
'at least 2', 'at least 3' and '4'"(pm)
- "generalization relations are useful for all kinds of inferencing,
including logical inferencing"(pm)
- ("not all expressions using complex KR constructs can be defined in an
ontology"
example: "there are an infinite number of possible
numerical quantification expressions"(pm)
)(pm)
- "even if all expressions using complex KR constructs could be defined in
ontologies, different users would use KR construct categories from
different ontologies and hence, when these categories from different
ontologies are not connected by identity/equivalence relations,
knowledge comparison/retrieval/sharing is reduced"(pm);
"the more expressive the KRL, the better for knowledge re-use";
argument: "the less expressive the KRL, the more a user may have to
represent knowledge in an imprecise and ad-hoc way"(pm),
objection: "the more expressive the knowledge, the more difficult
it will be to use for efficient reasoning"
(fg, objection: "applications requiring precision need the expressivity"(pm),
objection: "applications requiring efficiency and not a great precision
can ignore some certain elements of the statements and, for
example, use simple graph-matching techniques"(pm)
);
"most practical_applications do not require a language as restricted as OWL"
specialization:
("most practical_applications requires a language more expressive than OWL"
argument: - "a practical_application requires the use_of_rules"(gpZarri)
- "the use_of_rules is hard with a DL"(gpZarri),
argument: "any notation for OWL can only be ugly and verbose"(gpZarri),
argument: "OWL is difficult for an average programmer to deal with"(gpZarri),
argument: ("OWL has been a flop from a concrete/commercial viewpoint until 2007"
argument: "THIS was acknowledged by the IEEE Intelligent Systems issue
of September/October of 2007"(gpZarri)
)(gpZarri),
argument: ` "DL-Safe-SWRL is decidable" has for consequence
"SWRL variables can only be bound to known individuals in a KB" '(gpZarri),
argument: ("expert systems of the seventies were more interesting than OWL"
argument: "in 2007 the best way to use rules along with a DL-language is
to re-use an expert system such as Jess or Algernon"(gpZarri)
)(gpZarri)
)(pm),
argument: "most programmers do not care about computational_tractability"(gpZarri),
argument: ("RDF is probably more useful than OWL from an 'applicability' viewpoint"
argument: ("several commercials_products use RDF and not OWL"
example: "Oracle_11g_RDF_database uses RDF and not OWL"(gpZarri),
example: "GroupMe! and not OWL"(gpZarri)
)(gpZarri),
)(gpZarri);
2.3. Mandatory tasks of the SUO?
Warning: the sentences in this sub-section are not yet normalized.
"the SUO endeavour includes setting up a better alternative to OWL"(gpZarri)
argument: "most practical_applications requires a language more expressive
than OWL"(gpZarri);
"the SUO endeavour requires more than informal_on-line_discussions"
argument: - "the SUO endeavour requires at least one
building_of_a_significant_application"(gpZarri)
- "the building_of_an_application requires more than
informal_on-line_discussions"(gpZarri)
- "the building_of_significant_application requires money"(gpZarri)
- "the building_of_significant_application reuires to meet each other
face to face"(gpZarri);