论文部分内容阅读
传统的命名实体识别方法直接依靠大量的人工特征和专门的领域知识，解决了监督学习语料不足的问题，但设计人工特征和获取领域知识的代价昂贵。针对该问题，提出一种基于BLSTM ( Bidirectional Long Short-Term Memory)的神经网络结构的命名实体识别方法。该方法不再直接依赖于人工特征和领域知识，而是利用基于上下文的词向量和基于字的词向量，前者表达命名实体的上下文信息，后者表达构成命名实体的前缀、后缀和领域信息;同时，利用标注序列中标签之间的相关性对BLSTM的代价函数进行约束，并将领域知识嵌入模型的代价函数中，进一步增强模型的识别能力。实验表明，所提方法的识别效果优于传统方法。
The traditional named entity recognition method solves the problem of lack of corpus of supervised learning by directly relying on a large number of artificial features and specialized domain knowledge. However, it is expensive to design artificial features and acquire domain knowledge. In order to solve this problem, a named entity recognition method based on BLSTM neural network structure is proposed. Instead of directly relying on human features and domain knowledge, the method uses context-based word vectors and word-based word vectors, the former expressing the contextual information of named entities that express the prefix, suffix, and domain information that make up the named entity; At the same time, the cost function of BLSTM is constrained by the correlation between tags in label sequence, and the domain knowledge is embedded in the cost function of the model to further enhance the recognition ability of the model. Experiments show that the proposed method is better than the traditional method.