Development of intellectual search system for scientific social networks

The semantic model of the scientific social network that allows to calculate weights of six kinds of links between its objects: "term-term", "author-author", "document-document", "author-document", "author-term", "term-document" is presented. The identification of the links allows to give the users advices on the objects that may be of interest to them, and to carry out an iterative search of objects. To do this, it’s necessary to calculate the matrixes reflecting the degree of similarity between the individual documents, terms and authors. It is proposed to use the Random Indexing method to calculate the matrixes. The representation of the semantic model as a graph allows to search iteratively in each of the above-mentioned modes as the traversing of the graph. Among the existing traversing algorithms the wave algorithm has been chosen. The elements of the search query refinement algorithm are used to implement interactive search in the Web application, which operates the proposed semantic model for social network analysis. To test this application the Russian-speaking segment of the Mendeley.com network and Russian network SciPeople.ru are indexed. As a result the semantic models of these networks are automatically built.