Decision Theoretic Fusion Framework for Actionability Using Data Mining on an Embedded System - Data Mining: Theory, Methodology, Techniques, and Applications - page 96

()

(7)

H O = h ( O / M )

3 3 3 3 3

where h 3 (O 3 ) is a speech synthesizer function to read the sentences.
J(Y) is a selecting function to choose a speech interaction tool. The currently se-
lected speech module is just enabled. The variable, M i is a given specific domain
knowledge. M 1 is an acoustic model to recognize the word, M 2 is not used and M 3 is
TTS DB. The variable, P is procedural knowledge to provide a user-friendly service
such as a helper function. I(P) is a function to guide the service scenario according to
the results of the speech interaction tool.
As a result, Z is an action to be performed sequentially. The final decision-making,
Z(t) represents the user's history to be processed when the decision is stored for a long
period of time. This can provide the statistical information when the user frequently
utilizes a specific function.
Sensory fused rules and data fusion rules can be a fusion framework for generation
of gathered information. In addition, extension of application service and integration
can be easily employed based on this framework. From this information, the genera-
tion method of user action statistics is described in next subsection.

3.3 Active Interaction Rule Generation

Users may interact at various service stages and domain knowledge may be used in
the form of a higher-level specification of the model, or at a more detailed level. In
our system, the speech interactive agent interacts with users using a conversational
tool. This user interaction information is applied to data mining which is inherently
an interactive and iterative process. This is due to the fact that the user has repeated
patterns that he or she frequently uses on specific applications with the car navigation
system. By using this information, the speech interactive agent asks the user whether
the user wants to perform a specific task, which is the statistical information to be
stored and estimated for a period of time according to the procedure in Figure 3. In
addition, the speech interactive agent can start a music player automatically according
to the days' weather broadcasts if the system has not been used for a long time. This
function can be set on or off manually on an application by a user. To obtain some
information for specific tasks, the speech interactive agent downloads and updates the
mined data from the data mining server via the wireless internet.

Fig. 3. Active interaction procedure using the user's frequency rule