基于长短时神经网络的古汉语分词系统

Ancient chinese word segmentation system based on long and short time neural network

在自然语言处理领域,分词是非拉丁语系语言处理的首要任务。而在中文自然语言处理中,常见的是针对现代汉语进行分词处理,对古汉语涉及得较少。针对这一问题,设计针对古汉语的分词系统。系统采用流行的深度学习方法,对中文首先进行分词,采用长短时神经网络LSTM提取序列特征;之后采用Softmax进行分类,设计针对古汉语分词的长短时神经网络。在测试中取得了理想的预期,基本可以满足设计需求。该系统一方面提高了教学的效率,降低了古汉语断句的难点;另一方面,在一定程度上提升了学习的兴趣。

In the field of natural language processing,participles are the primary task of non-Latin language processing.In Chinese natural language processing,it is common to classify words for modem Chinese,and there are very few related to ancient Chinese.In response to this problem,the word segmentation system is designed for ancient Chinese.The popular deep learning method is used in the system which firstly cut sentence and then feed the sequence to LSTM to design long and short time neural networks for ancient Chinese word segmentation.It meet the design requirements.On the one hand,it improved the efficiency of teaching,lowered the difficulty of the sentence in ancient Chinese,and on the other hand,the interest of learning is raised at certain extent.