Information Technology Reference
In-Depth Information
which the parameters of each rule belong to the
document structural contexts and selects the rule
with the higher degree. If the difference between
the degrees of the best solutions is negligible,
the analyzer decides that it is unable to make a
decision and rejects the element. We do not pres-
ent this rule selection component more in detail,
because we believe that it is outside the scope of
this chapter.
Once interpreted, the new element is created;
it replaces the parameters of the applied rule in
the document. New structural contexts are cre-
ated to help interpreting the following elements;
this is modelled by the DCC block. Once a rule
is applied, the current iteration of the analysis
process is finished. Then, a new iteration begins,
in order to check if this new element interacts
with other elements of the document to constitute
a more complex symbol: we try to eventually
apply a rule on the new element, and so on until
stability; as a consequence, a stroke can imply
a sequence of transformations. If no rule can be
applied on a stroke (i.e., if the first iteration does
not succeed), it is rejected and disappears from
the editing window.
In order to explicit this interpretation process,
Figure 9 presents its mechanism on one particular
example, which is the interpretation of a sharp.
In this example, the user first draws a vertical
segment on the left of a filled-note head (1). This
hand-drawn stroke is recognized by the system,
and replaced by its neatly retranscribed symbol.
This segment does not, for now, interact with other
existing elements of the document to from a more
complex one: the analysis process is over. The
mechanism is the same when the user adds two
horizontal segments on the left of this note head
(2 and 3). Finally, the user draws a stroke which
is supposed to be the second vertical segment that
ends the sharp symbol. In the first iteration of the
interpretation process, the stroke is interpreted as
a vertical segment (4). This new element is then
analyzed in the second iteration: it interacts with
the three other segments to constitute a sharp (5).
As this new symbol does not interact with other
existing elements of the document to constitute
a more complex symbol, the analysis process is
over. We would like to remind the fact that the
order in which these four segments are drawn is
not constrained.
User Validation
As the recognition process is eager, the result of
the analysis is displayed directly as the user is
drawing. We can then exploit the human-com-
puter interaction and integrate the user in the
interpretation process to validate or reject the
results. Thus, if after the display of the answer,
the user goes on with the drawing of the docu-
ment, he implicitly accepts and validates it; on
the contrary, if he does not agree, he can delete
the new element with a deletion gesture and so
explicitly reject it. The main consequence of this
process is that it is not necessary for the analyzer
to question a decision made beforehand because
it has been validated by the user. We believe that
it is pertinent because it could be perturbing for
the user to see the interpretation of an element
changing after drawing another one. It is a major
Figure 10. Three different visualization modes: novice mode (left), expert mode (middle), and contextual mode
(right)
Search WWH ::
Custom Search