摘要： 提出了基于前后文n-gram模型的古汉语句子切分算法，该算法能够在数据稀疏的情况下，通过收集上下文信息，对切分位置进行比较准确的预测，从而较好地处理小规模训练语料的情况，降低数据稀疏对切分准确率的影响。采用《论语》对所提出的算法进行了句子切分实验，达到了81%的召回率和52%的准确率。

关键词: n-gram模型, 数据稀疏, 平滑技术, 基于前后文的n-gram模型

Abstract: An algorithm of punctuating the sentences in archaic Chinese language based on context n-gram model is proposed in the paper. The algorithm can make comparatively accurate prediction of the punctuating-positions of the text under data-sparse instances by collecting and calculating context information to better analyze small-scaled corpus and meanwhile, to bring down the effects of the data-sparse plight on the global accuracy. At last, the paper selects the analects of Confucius ( Lunyu ) to test the algorithm introduced, and the results show that the recall and the precision achieve 81% and 52% respectively.

Key words: N-gram model, Data sparse, Smoothing technology, N-gram model based on context