An Improved Method of Reliable Content Extraction from Partially Unreliable Publications in Plural Number of Sources over the Network

Abstract

This paper discusses a method of simplifying the model of information dissemination and mainly eliminating errors statistically without going into semantics for the same content collected from many information sources scattered over the network. In the methods we proposed in the past, it is assumed that the same content collected from the sources can be divided into parts that have independent values. By taking a majority vote about the value of each part, the content consisting of only the part with major values is reconstructed. For the improvement of the extraction accuracy of the reliable part of the target content, more efficient weighting ways of majority voting and their effects are discussed.