Information search model based on the use of percolation properties of semantic networks of texts

The paper examines the models of information search in the texts presented in the multidimensional vector space. Describes approaches to semantic representation of a text document. The concept of information conductivity of the document which can be used for the information search task is discussed. The developed model of construction of semantic multiconnected document and offered algorithm of construction such network for the document from the collection of texts. The second part of the article describes the application of percolation theory for the description of information conductivity of multiconnected semantic networks. Since the percolation threshold determines the loss of meaning and the separation of the text into separate unrelated fragments, its value can be taken in determining the relevance of documents for solving problems of information search. Shown the correlation of percolation threshold and "semantic force" of the document. The last part of article describes the "greedy" algorithm of clustering documents using the value of percolation threshold as measures of information conductivity. In conclusion there is given approaches for performing practical calculations based on the described theoretical approaches.