Soft clustering method for text mining, with an opportunity to attribute them to different semantic groups

The work describes new soft clustering method for text mining, developed by the authors, with an opportunity to attribute them to different semantic groups. The work of this algorithm is based on usage of definition parameter, when deterimining clustering accuracy, which can be defined for texts of various subject areas either on basis of percolation properties of text clusters, or estimated theoretically based on the redundancy model for text messages. By the author’s assumption, proposed algorithm must have speed and accuracy for clustering of texts, depending on value of clustering accuracy parameter. The advantage of this algorithm is that no need to set an initial value of clusters.