Stochastic language modeling using high linguistic information

首先，我们提出了一种逐点分词的方法，并实现了分词的改进。然后，我们创建了一个由词典例句和带有依存信息标注的报纸文章组成的语料库。我们还提出了从随机切分或标注的语料库中进行随机标注和语言建模的方法。

First we proposed a pointwise method and realized an improvement of word segmentation. Then we created a corpus consisting of dictionary example sentences and newspaper articles annotated with dependency information. We also proposed stochastic annotation and language model building from a stochastically segmented or tagged corpus.