Searching Social Networks Sample Clauses

Searching Social Networks via Referrals‌ ▇▇ and ▇▇▇▇▇ [315] focus on using referrals for searching dynamic social networks. However, no social network analysis has been used in the peer selection process. The authors argue that because building and maintaining a peer’s social network is not feasible, distributed search through referrals is more promising. In a referral system, peers send queries to other peers of the selected contacts. A response can either be an answer or another referral, allowing referrals to propagate in the social network. But how do peers select which peers to send their query to? To achieve that, each peer maintains a profile for itself and an acquaintance model for each of its acquaintances, modelled via the vector space model (VSM). Similarly, a query is modelled as a term vector. The similarity between a 68 October 5, 2009 LiquidPub/2009/D1.1/v2.0 query and expertise of a given peer is defined as the cosine of the angle between them. This similarity measure is then used by a peer to help it select other peers to query. In addition to the similarity measure, a peer also considers the sociability of others (or their ability to give good referrals). Weighted referral graphs are used to help the peer decide which querying peer to follow first. When another referral is received, the peer might decide to pursue it, even if the referred peer is not in its acquaintance list. This is how acquaintances are added. But if an answer is received, the peer updates its acquaintances by assigning rewards and penalties. Both the expertise and sociability vectors of the peer who sent the reply will be updated according to well defined functions. Note that peers are assumed to be trustworthy is answering queries. For instance it is assumed that a peer answers a query only if it is confident of its expertise, or it may send back another referral only if it is confident in the relevance of the peer being referred.