Cooperative game theory methods for text ranking

Abstract

A method of ranking the corpus of texts of a news portal, based on measures of graph centrality, is proposed. Each text is assigned a vertex of a certain graph, and its structure is determined based on the semantic connectivity of the texts. As a measure of centrality, the Myerson value is used in a cooperative game on a graph, where the number of simple paths in a subgraph of a certain length m is chosen as a characteristic function For different values of m, the ranking based on the Myerson value will be different. For the final ranking, it is proposed to use the ranking procedure based on the tournament matrix. The operation of the ranking algorithm is illustrated by numerical examples related to a specific news porta.