The Social Spiders in the Clustering of Texts

Abstract

In this paper the authors experiment and test a new biomimetic approach based on social spiders to solve a combinatorial problem ie the automatic classification of texts because a very large data stream flows and particularly on the web. Representation of textual data was performed by a method independent of the language ie n-gram characters and words because there is currently no method of learning that can directly represent unstructured data (text). To validate the classification, the authors used a measure of evaluation based on recall and precision (F-measure). During the experiment, the authors found a powerful visualization tool in social spiders that they exploit to make visual classification.