RANKING OF HETEROGENEOUS INFORMATION OBJECTS

摘要

A Computer implemented method of ranking information objects within a collection of information objects linked by a collection of links, the collection of information objects comprising information objects of a first nature and of a second nature, the links being associated to a link type selected among a plurality of link types, the method comprising: allocating a qualification weight to each link, selecting a plurality of paths comprising a sequence of information objects linked by a sequence of the links, for each information object, computing a score as a function of the respective contributions of the links that point to the information object, the contribution of a link being a function of the number of times the link has been selected in the path selection step and the qualification weight of the link, ranking the information objects as a function of the respective scores.

主权项

1. A Computer implemented method of ranking information objects within a collection of information objects stored in data repositories and linked by a collection of links, the links being oriented, the collection of information objects comprising information objects of a first nature and information objects of a second nature, the links of said collection of links each being associated to a link type selected among a plurality of link types, the method comprising a step of ranking the information objects of the first nature as a function of the respective scores of the information objects of the first nature, the method wherein comprises:
allocating a qualification weight to each link, the allocated weight being defined as a function of the link type associated to the link, selecting a plurality of paths within the collection of information objects, each path comprising a sequence of information objects linked by a corresponding sequence of the links, wherein each successive link of a path is selected randomly among the links that originate from a same information object using link selection probabilities, and wherein the selecting of a path further comprises a step of terminating the path, the step of terminating the path being made randomly based on a damping factor, for each information object of the first nature, computing a score of the information object as a function of the respective contributions of the links that point to the information object, the contribution of a link being a function of the number of times the link has been selected in the path selection step and the qualification weight of the link.