Essay: Semantic Routing
Essay: Semantic Routing
A method of routing which is said to be more focused on the nature of query than on network topology is Semantic routing. It improves on previous methods of traditional routing by setting the priority of the nodes on the basis of how good the nodes have provided information about the content of the query previously.
In Semantic routing, the data needs to have a semantic description attached to it, in order to be eligible to be searched. A popular solution identified by Oram (2001) is to use RDF meta-data. By tagging the data/files with RDF meta-data enable a semantic web to be constructed in a P2P manner. Semantic routing is different from traditional routing techniques because prospective nodes are selected on the basis of confidence that other nodes have on it for responding correctly to the incoming query regardless of their position in the network.
Tempich, Staab, and Wranik (2004) in their paper have mentioned “Remindin” which is based on a semantic routing algorithm and was developed with the goal of mimicking social networks. In addition to this, a ranking of peers in semantically routed networks is also done through collaborative filtering.
In a Semantically routed network, every time a node responds to the query of its peers correctly; its peers adjust their confidence on the node regarding the type of the query. The information that is associated with the type of query relies greatly on the type of semantic data that the network is dealing and the strictness of the peer confidence rating algorithm. Tempich, Staab, and Wranik (2004) have shown that automated relaxation of queries can also result in more robust searches.
Kumar (2006, pp.703) states, that an important factor that needs to be incorporated in routing algorithm for it to be effective in the long term is persistence. Each node must maintain a constant identifier within the network in order to maintain its confidence rating. During the initial stages of network forming, none of the peers have ratings for any node and search results in returning of random nodes, as it happens in the traditionally routed network. Tempich, Staab, and Wranik (2004) have observed that returning of random responses to a certain amount of all requests prevents the ‘Overfitting’ effect, which denotes a state when the confidence data associated with the nodes becomes inflexible and rigid.