AN EVALUATION BASED ALGORITHM FOR TAGGING POS OF CHINESE CORPUS WITH A PURE PROBABILISTIC MODEL

An evaluation based algorithm is proposed to tag the part of speech （POS） of Chinese words with a pure probabilistic model.In the algorithm, the “confidence intervals” in the estimation of parameters is employed to construct a dynamic evaluation function. The test reveals that the algorithm can guarantee its defined high accuracy, as well as a definite applicability. When its precision is 94.8%,the recall （applicability） can reach 81.5%.