The core of this work is to realise a system of classification for Arabic texts (SCAT) based on the inter-textual distance theory for Arabic language. This theory assumes the classification of texts according to criteria of lexical statistics, and it is based on the lexical connection approach. Our objective is to integrate this theory as a tool of classification of texts in Arabic language. It requires the integration of a metrics for the classification of texts using a database of lemmatised and identified corpus which can be considered as a literature reference for times, kinds, literary themes and authors and this in order to permit the classification of anonymous texts.