基于BI-GRU-CRF模型的中文分词法

Chinese Word Segmentation Based on Bi-directional GRU-CRF Model

循环神经网络作为一种处理时序数据的有效模型,已在序列标注问题上得到了广泛应用。为解决序列标注中典型的中文分词任务,基于门限循环单元(Gated Recurrent Unit,GRU)神经网络,提出了一种改进的双向门限循环单元条件随机场(BI-GRU-CRF)模型,该模型不仅可以通过双向门限循环单元有效利用双向上下文信息,而且可以通过条件随机场层联合考虑相邻标签间的相关性,得到全局最优的标记序列结果。在常用的中文分词测评集(PKU、MSRA)以及由构建的军事领域分词语料上,分别采用四词位及六词位标注法进行了实验,结果表明BI-GRU-CRF模型具有良好的分词性能,且六词位标注法可以改进分词效果。

As an effective model for processing time series data,recurrent neural network has been widely used in the problem of sequence tagging task. In order to solve the typical sequence tagging task of Chinese word segmentation,this paper proposes an improved bi-directional gated recurrent unit conditional random field(BI-GRU-CRF) model based on the gated recurrent unit(GRU) neural network. This model can not only effectively utilize two-way contextual information through bi-directional gated recurrent units,but also get the globally optimal tagging sequence as a result by considering the correlation between adjacent tags through conditional random field. In this paper,experiments are carried out on the common Chinese word segmentation evaluation set(PKU,MSRA)and the segment corpus in the military field constructed in this paper with the four-tag-set and sixtag-set respectively. The results show that the BI-GRU-CRF model has high performance in Chinese word segmentation,and the six-tag-set can improve the effect of word segmentation.