EpidemicSurveillanceaimsatdetectingdiseaseoutburstsintheworld in order to provide useful information to health authorities. Many automatic systems have been conceived to help these authorities to mine the available data with a special focus on press articles. The main goal of these systems is to select relevant articles by means of text classification. The secondary goal is to extract valuable information from these relevant texts . One of the main challenge is to handle many languages with different properties, various availability of language resources (lexicons, POS taggers...) and annotated data. In this paper we present a state of the art on Text Classification as well as Information Extraction for the Arabic language and we test different options for designing a lightweight system to process texts in written Arabic. We show that Arabic language has particular properties, making it difficult to handle properly without improving existing approaches. We propose improvements of an existing lightweight approach that would be promising for Arabic as well as more poorly endowed languages.