【摘要】 为减轻人工标注训练语料库面临的瓶颈问题，提出了一种基于 Ｎ Ａ 假设带标训练语料库的自动构造方法·为了检验该方法的有效性，将自动获取的带标训练语料库用于词性标注应用中，２ 万词次的开放性测试结果的准确率为９３ ．１ ％ ，其中词性兼类消歧准确率为７９ ．３ ％ ，未登录词词性确定准确率为８８ ％ ·更多还原

【Abstract】 An approach to training data acquisition based on NA assumption from a raw corpus was presented. Using the training data trains parameters of the grammatical tagging model. The precision of grammatical tagging on the open test set is 93%. The precision of multi POS′s word disambiguation is 79 3%. The precision of determining grammatical category of unknown word is 88%.更多还原