Information Technology Reference
In-Depth Information
is therefore in the Beaver category. Practically, this type of change is managed at the
level of the modeling and more specifically at the level of the business processes in
which it is possible to carry out a category change following a user action.
- The change can also occur in an uncertain manner, based on contextual
information that the application and the platform are able to recuperate and manage.
Thus, if the user is no longer using the application and if a sensor means that we
know he is talking, we can presume that he is in the Peacock category. Similarly, if
the user is no longer using the application and a sensor determines he has not spoken
nor moved for more than five minutes, we can presume that he is in the Marmot
category. This type of adaptation is close to the works of [SIE 03] that, basing
themselves on information obtained from sensors, propose carrying out a dynamic
contextual adaptation according to the modifications of the exterior environment and
the physiological states of the user.
If the change according to the actions of the user is done systematically in a sure
way, the change in the function of contextual information occurs following an
uncertain methodology based a certain number of heuristics. In any case, the
automatic change of category is done according to probabilistic methods that are
based on a certain number of criteria that enable us to define, for sure, whether or
not they belong to a specific category. For this to happen, the criteria must be
measurable and quantifiable and in particular must be fixed in the context of the
execution of each application. Thus, if we take our example of the duration of non-
use of the application as a criterion for category change, we have:
- duration of non-use >2 min = possibility of changing category to Peacock ;
- duration of non-use >5 min = possibility of changing category to Marmot .
This simple example shows a limitation of the approach in the sense that, if we
only use these criteria, when the duration of non-use is greater than 5 min it is
impossible to define which category the user is in ( Peacock or Marmot ?).
To limit error risks during category changes, for each business process we
propose to associate a confidence level with regards to the different categories. This
confidence level can take on a value between 0 (the user is definitely not in this
category) and 1 (the user is definitely in this category). For example, when the user
consults the list of the latest films that have just come out at the cinema, we can in
principle define with certainty that he is not in the Beaver category (confidence level
fixed at 0.0) but more probably in the Peacock category (confidence level fixed at
least at 0.5). In the event where a degree of confidence is not explicitly associated
with a category, it is considered as being equal to 0.0.
Search WWH ::




Custom Search