Information Technology Reference
In-Depth Information
three components: The dialog situation, the agent role, and the conveyed functional role.
To illustrate this, let us detail the main FR types used in the our tutored learning dialog. Our conven-
tion for FR naming is: '
verbal
predicate
'-'
object
type
'('
argument
').
1.
give
-
knowledge
(): Used to teach a knowledge and introduce an exchange.
Argument's general form: (
P
→
Q
) or (P↔Q).
Example:
give
-
knowledge
(
cat
(
x
))→
mortal
(
x
)); “Cats are mortal.”
When uttered by the 'teacher', a
give
-
knowledge
() argument has to be evaluated as true by the
'student' (see 'FR Interpretation Axioms' section).
2.
askfor
/
give
-
information
() (boolean evaluation case):
•
askfor
-
information
().
Argument's general form: Either (
P
), or (
P
→
Q
), or (P↔Q), with or without variable.
Examples:
askfor
-
information
(
cat
(
F
olley
)); “Is Folley a cat? ”
askfor
-
information
(
cat
(
x
)→
mortal
(
x
)); “Are cats mortal ? ”
When conveyed to the interlocutor this function bids him/her to answer. In a very cooperative
framework as the one we need to install between artificial agents, the 'teacher' agent is compelled to
answer with a
give
-
information
() utterance, which gives the interpretation of the formula (
cat
(
x
)→
mortal
(
x
)) according to the 'teacher'.
•
give
-
information
().
Argument's general form: Either (
True
), or (
False
), or (
Unknown
).
Example:
give
-
information
(
t
rue
): “Yes.”
3.
give
-
explanation
() (predicate case).
Argument's general form: Either (p(x)↔P), or (
Unknown
).
Example:
give
-
explanation
(
cat
(
x
)↔(
animal
(
x
)∧
pet
(
x
))): “A cat is a pet animal.”
A
give
-
explanation
() formula is provided as an answer to a question of the type: “what is X ? ” or
“Why/how is X related to Y ? ”. In other words, when a 'student' has no value to a predicate, or
cannot relate it to another, the 'student' asks the 'teacher' to provide the links between the unknown
element and other possibly known predicates. The situation can be triggered by an
askfor
-
expla
-
nation
() clause taking as an argument the unknown predicate or formula. The 'student' expects
the 'teacher' to provide a formula in which known predicates are related to the unknown one. By
this process, the 'student' might augment its KB while increasing its connexity. A “Why/how is X
related to Y ? ” question is about connexity, and a “what is X ? ” question increases KB elements
through KB connexity.
4.
say
-
dis
/
satisfaction
: Tells the other agent that the last provided data has (has not) been well under-
stood. This is a meta-evaluation clause, since it has no direct argument, but leads to the evaluation
of the interaction (and not of the formula).
say
-
dis
/
satisfaction
is particular to dialog modeling (most
linguistic and psycholinguistic theories account for interaction evaluation), and is very useful in
checking dialog feedback.
There are some FR we do not detail here (
askfor
-
knowledge
,
askfor
/
give
-
example
,
askfor
/
give
-
precision
,
askfor
/
give
-
reformulation
) likewise some specific uses like the type
askfor
/
give
-
informa
-
tion
in the case of an evaluation by a function. So FR are dialogic clauses leading to the interpretation
of exchanged formulas. A functional role of the '
askfor
' kind implies one or a series of clauses of the
Search WWH ::
Custom Search