Design Considerations for a Social Network-Based Recommendation System (SNRS) - Community-Built Databases: Research and Development - page 96

Fig. 4.7 The student social
network (a) before and (b)
after semantic filtering. Each
node represents each student.
For a pair of students U and V ,
node U has a directed edge to
node V if U rates V with a
value of 3 on at least one of
the three factors and U rates V
with a value of 3 for at least
two of the three factors
U considers student V as a friend, then there is a corresponding directed edge from
U to V .
SNRS with semantic filtering ( SNRS-SF ). In this method, when predicting fine-
grained user ratings, we consider student V as student U' s friend if U rates V with a
value of 3 on the given factor. When predicting overall ratings, we select V as U' s
friend if U rates V with a value of 3 on at least two of the three factors. Figure 4.7b
shows the social network of the students after we apply semantic filtering to overall
ratings. When compared to Fig. 4.7a , we can see that many social relationships have