Chapter 6 Viewpoint Focus Modelling
6.6 Comparison of Viewpoint Focus Models
6.6.3 Example Viewpoint Focus Models Comparison
The modelling properties inherited from FCA over the ontologies provide quantitative (structural) and qualitative (regional) indicators for diversity between two or more viewpoint focus models.
Let us consider two viewpoints and on a set of digital objects , where:
Figure 6.18 depicts the contrastive semantic map for two sets of annotated ontology entities in the mental state branch of the WNAffect taxonomy from the simulator dataset.
Figure 6.18 The contrastive semantic map of the annotated ontology
entities sets and .
Given the two sets and using conventional set operations one can identify that:
has more entities annotated than , ( ); has 8 common ( ) entities with ;
has more distinct ontology entities than compared to each other (differences of sets).
ViewS complements the comparison metrics by extracting, representing and comparing the viewpoint focus models from the annotation sets to identify what are the similarities and differences between the viewpoint focus.
Structural Comparison. The two viewpoint focus models for these annotation sets are depicted by the lattices in Figure 6.19 using a semantic distance
threshold 3. The lattices indicate that is broader than as more main focus elements (see Section 6.5) are extracted ( :12, :10). This observation does not necessarily reflect the fact that includes more unique entities than , as if these were aggregated together based on the distance threshold would result to fewer main focus elements. This comparison is enabled with ViewS as the focus models explicitly denote the difference. From the focus models it is also observed that contains more focus elements in total than , organised in more layers. This indicates that more implications exist between the focus elements in , therefore closer aggregates are derived than from . Although appears broader, seems more condensed considering the semantic space, again with respect to the application of the same distance threshold.
layers: 5, elements: 22, main elements: 12
layers: 7, elements: 27, main elements: 10
Figure 6.19 The viewpoint focus models derived from the annotated
ontology entities sets.
Differences are observed in the structure characteristics. appears broader than , as more main focus elements are extracted. However, appears more complex, as more elements occur in the lattice organised in more layers.
Regional Comparison. For the example focus models the corresponding cross-table for the comparison of the extracted main focus elements includes 120 pairs ( ), including: 3 equal, 2 includes, 12 overlap and 103 disconnected. Example illustrations of the qualitative comparison relations are shown in Figure 6.20 extracted with ViewS-Microscope. Note that in this example equality is not very helpful as it only concerns a single ontology entity in the focus elements.
Figure 6.20 Example qualitative comparison relations for selected focus
elements from two viewpoint focus models on the mental-state WNAffect taxonomy branch.
Using the comparison cross-table each focus element from a focus model (by row or similarly by column) can be examined across the focus elements of the other model. From such examination, conclusions can be drawn regarding which focus elements appear more equal, disconnected, inclusive, included and overlapping, as well as an overview of the similarities and differences. In the previous example, observing the contrastive semantic map (see Figure 6.18) the two viewpoint focus models appear very
overlapping, especially around negative emotion. Indeed, all the 12 overlapping pairs of focus elements relate to this ontology branch (see Figure 6.21 below), between 3 main focus elements from the first model and 4 from the second respectively.
Using the ViewS viewpoint focus modelling presented in this Section, the UGC from the simulated environment can be examined by over viewing and comparing different focus models. The next Section illustrates the application of ViewS on the same content and setup used in the study with the simulator designers to validate their observations using the computational instruments of the focus modelling approach.
Figure 6.21 A sample of 4 pairs out of the 12 overlapping focus elements
between the two focus models.
ViewS enables cross-table comparison to identify relations between focus elements and understand similarities and differences.
6.7 Discussion
In this chapter the viewpoint focus modelling with ViewS was presented. ViewS adapted the FCA computational framework using as input:
ontologies to represent domain knowledge; and,
semantically annotated data sets (which linked UGC to ontology entities);
and produced as output:
semantic relations of the annotated ontology entities represented as formal contexts;
viewpoint focus elements represented as semantic clusters and aggregates based on the derived relations; and,
viewpoint focus models represented as formal concept (focus element) lattices for different ontology branches.
For comparison of viewpoint focus models RCC was exploited in a simplified version to include equality, inclusion, overlap and disconnection. Focus extraction and comparison were illustrated with an example. Particularly for comparison, two approaches were discussed: structural – which concerns the lattice structural characteristics, and regional – using the RCC on the main focus elements.
Using the main framework components i.e. viewpoint context (formal context), focus element(formal concept), focus model (concept lattice) and focus model comparison, several observation can be made to evaluate the underlying modelling assumptions. A reflection is following on each component of the model discussing strengths and limitations, as well as indication for future extension.