Chinese full text information retrivial is one of the efficient technologicies of information retrivial,and a lot of researches have been made on the application of the maximum entropy modeling in the natural language processing during recent years. This paper presents a retrivial system based on maximum entropy principle. The context of speech tagging and word frequencytagging included in database are considered,and Chinese word segmentation algorithm is put forward.By adopting object-oriented and the model driven programming method the algorithm can be used in a system design.