Abstract: Lexicon based feature extraction is an effective method in Chinese information processing. But it highly depends on the lexicon used. The coverage of a lexicon determines the correctness of word segmentation. Therefore, it is crucial to learn new words continuously and to update lexicon dynamically, making the whole Chinese information processing system more adaptive and dynamic. Proposed in this paper are an innovative theoretical model and the implementation model for dynamic information updating in lexicon based search engine. The results of testing show that this model can reduce the time for re establishing the information database and can greatly improve the precision of a search engine.