摘要：To face what is called the Information dump, information retrieval systems have been used, in order to help people to organize and classify knowledge, and to retrieve pertinent information rapidly when it is needed. In this aim, we have tried to improve the performance of the existing Arabic information retrieval systems by using, in addition to some natural language processing, the latent semantic analysis method. This later consists in analyzing relationships between the terms that are contained in a set of documents and producing concepts related to the documents and terms. Moreover, we have included to this system and tried some weighing schemes that have proved their performance, specially the Okapi BM-25 scheme. Regarding the Amazigh language, in the aim to compensate the lack of an Amazigh information retrieval system, especially for the tifinagh script, we have proposed a new information retrieval system structure that is in the process of evaluation.
