【摘要】 论文在传统一阶隐马尔可夫模型的基础上，针对隐马尔可夫模型结构信息挖掘不全面的问题，提出了一种双层隐马尔可夫模型。双层隐马尔可夫模型在使用Baum-Welch算法的过程中将词性序列视为观测序列，通过Baum-Welch算法提取更多信息并最大化词性序列概率从而更加贴合实际情况，同时对Viterbi算法做了相应的改动。模型在Penn Treebank语料库和Groningen Meaning Bank语料库上进行10折交叉验证，并与传统一阶、二阶隐马尔可夫模型进行对比。结果表明双层隐马尔可夫模型相较传统一阶、二阶隐马尔可夫模型词性标注正确率更高。更多还原

【Abstract】 Based on the traditional first-order hidden Markov model,this paper proposes a double-layer hidden Markov model to solve the problem of incomplete structural information mining of hidden Markov model. In the process of using the Baum-Welch algorithm,the double-layer hidden Markov model regards the part-of-speech sequence as an observation sequence,and extracts more information and maximizes the probability of the part-of-speech sequence through the Baum-Welch algorithm,which is more suitable for the actual situation. made corresponding changes. The model is cross-validated with 10 folds on the Penn Treebank corpus and the Groningen Meaning Bank corpus,and compared with traditional first-order and second-order hidden Markov models. The results show that the double-layer hidden Markov model has a higher accuracy of part-of-speech tagging than the traditional first-order and second-order hidden Markov models.更多还原