Semantic Knowledge Representations for Soft Data Fusion - Efficient Decision Support Systems: Practice and Challenges from Current to Future - page 26

Fig. 16. Percentage of programs correctly fused and identified with different strategies
6.4 On the usefulness of association
The results of the fusion of TV programs that are scheduled on periods of low audience
are very bad. Among other particularities, we observed that the TV magazine has "holes",
especially for non-popular channels. During such periods, as next program to be broadcast,
the magazine source of information gives a program that will actually be broadcast several
hours after, whereas, the DVB gives the real next one.
Fig. 17. compatibility testing
The two descriptions are then incompatible and the resulting fused program is not well
formed (it has two different titles or begin dates for instance). To overcome such problems,
we introduced the use of the association phase.
Figure 17 shows the different percentages of program correctly found, first without
association, then using one based on title similarity and distance between begin and end times.
7. Conclusion and future work
We studied the issue of soft data fusion. The aim is to propose a generic approach and a
generic framework for high level information fusion. The information items represent the
descriptions of (part of) complex situations that themselves contain several actors or objects
linked together through semantic relationships.
Besides, the proposed framework enables fusing informations items coming from several
sources. Discrepancies among pieces of information are studied and we detect that two lightly
different pieces of information concern the description of the same situation, and then choose
what the fused description should look like.
We focused on three aspects regarding information fusion: the modeling of the situation of
interest, the association phase, which aims at deciding whether two observations concern