Database Reference
In-Depth Information
Unlike related work, this model has several new features. First, the social
information network includes new entities corresponding to users and social anno-
tations in addition to documents and author nodes presented in [ 18 , 50 ]. This helps
to estimate document relevance based on their social production and consuming
contexts. Second, we include citation links as social interactions between authors of
scientific papers enriching thus their mutual associations previously based on
coauthor relationships only [ 41 , 51 ]. Finally, we define a weighting model for
edges connecting social entities in the contrast of approaches presented in [ 41 , 50 ]
modeling bibliographic resources using a binary network model. Specifically,
weights are assigned to coauthorship, citation, and authorship edges to evaluate
influence, knowledge transfer, and shared interest between authors.
6.5 A Social Retrieval Model for Literature Access
This section presents our novel retrieval approach for literature access [ 52 ] based on
social network analysis. In fact, we investigate a social model where authors
represent the main entities and relationships are extracted from coauthor and citation
links. Moreover, we define a weighting model for social relationships which takes
into account the authors' positions in the social network and their mutual collabora-
tions. Assigned weights express influence, knowledge transfer, and shared interest
between authors. Furthermore, we estimate document relevance by combining the
document-query similarity and the document social importance derived from
corresponding authors. To evaluate the effectiveness of our model, we conduct a
series of experiments on a scientific document dataset that includes textual content
and social data extracted from the academic social network C ITE UL IKE .
6.5.1 The Social Information Retrieval Model
An information retrieval model is a theoretical support that aims at representing
documents and queries and measuring their similarity viewed as relevance. For-
mally and based on the representation introduced in [ 53 ], the social information
retrieval model can be represented by a quintuple [ D , Q , G , F , R ( q i , d j , G )], where D
is the set of documents, Q represents the set of queries, G is the social information
network, F represents the modeling process of documents and queries, and R ( q i , d j , G )
is the ranking function including various social relevance features and taking
into account the social information network topology G . This function can be
defined by combining the subset of the flowing factors: the topical relevance, the
social importance of actors, the social distance, the popularity, the freshness, and
the incoming links and tags [ 21 ]. The social information network G represents the
Search WWH ::




Custom Search