Social link recommendation by learning hidden topics

Abstract

In this paper, a new approach to predicting the structure of a social network without any prior knowledge from the social links is proposed. In absence of links among nodes, we assume there are other information resources associated with the nodes which are called node profiles. The task of link prediction and recommendation from text data is to learn similarities between the nodes and then translate pair-wise similarities into social links. In other words, the process is to convert a similarity matrix into an adjacency matrix. In this paper, an alternative approach is proposed. First, hidden topics of node profiles are learned using Latent Dirichlet Allocation. Then, by mapping node-topic and topic-topic relations, a new structure called semi-bipartite graph is generated which is slightly different from regular bipartite graph. Finally, by applying topological metrics such as Katz and short path scores to the new structure, we are able to rank and recommend relevant links to each node. The proposed technique is applied to several co-authorship networks. While most link prediction methods are low precision solutions, the proposed method performs effectively and offers high precision.

Who proposed the relationship?

“…Another kind of popular approach utilizes the random walk to measure the node proximity on the whole network [19,2]. Furthermore, matrix factorization methods [27,5], contentbased methods [8,25], attribute-based methods [39] and behavior modeling based methods [41,42,40] have been studied respectively. Supervised methods are also investigated [9,22].…”

Social Link Prediction in Online Social Tagging Systems

“…Makrehchi [2011] constructed a semi-bipartite graph of extracted hidden topics from user profiles and then applied topological metrics such as Katz [Katz 1953] and short path scores to rank and recommend users. Makrehchi [2011] showed that this method outperforms approaches that rely on similarity measures of feature vectors (i.e., Bag of Words) and low rank approximation (i.e., Latent Semantic Indexing (LSI)). We extend this approach by considering resources and metadata to represent users' interests instead of documents consisting of words.…”

“…Topic-actor matrix in particular represents a bipartite graph linking topics to actors. Using matrix , we can build a semi-bipartite graph G [Makrehchi 2011] …”

“…Makrehchi [2011] used Katz score [Katz 1953] to predict the missing values of the unknown block S, such that S = Katz(G), in an academic social network. Katz score, a generalization of degree centrality, measures the degree of influence of an actor in a social network [Katz 1953].…”

“…As the similarity threshold decreases, more links are added, leading to more true positives but also to more false positives. We determine the best threshold value automatically based on the probability of the existence of a link in a social network [Makrehchi 2011]. In a sparse, directed social network, the number of existing links is considerably smaller than all possible links.…”