摘要：In this work we submit the results obtained for the building of a statistical model of the Arabic language, adopting for a word the prefix*-stem-suffix structure based on the lattice. That solution allowed us to keep all the possibilities of word segmentation, which is one of the issues we have met when building the aforementioned model. The language has been evaluated from a corpus made up of 100K words and has been tested on a corpus of 7K words. The results and the analysis are submitted in this document
关键词：Automatic Speech Recognition; Language Model;Arabic Language; SRILM.
