Entities Extraction becomes very important for developing many applications of Natural Language Processing (NLP). In this paper, we present a new algorithm to extract entities from Arabic text. The approach uses the semi-structured knowledge source: Arabic Wikipedia to predict the words that constitutes an Arabic entity. Our method is generic and can be applied directly to other languages to extract entities. The proposed method has been designed to analyze Arabic text hierarchically with variable length N-gram. The experimental results have proven that the proposed system is very efficient in detecting entities from large set of Arabic news