CRF与规则相结合的军事命名实体识别研究

作战命令的分词是作战指令自动化生成、文图转换等各种指挥自动化技术的重要基础。在作战指令进行分词处理的过程中,军事命名实体的识别是技术难点之一。命名实体是信息的主要载体,它的识别是军事信息抽取的关键。提出了一种基于CRF模型与规则相结合的命名实体识别方法,结合基本特征与外部词典特征,提高了实体识别效率;在后期进行规则优化,最终实现高效的命名实体识别。实验证明,该方法是行之有效的。能够成功解决命名实体的自动识别问题。

Research on CRF and Rules based Military Named Entity Recognition

The segmentation of the command orders is one of the basics of C3I applications such as the auto-generation of command orders, automated documents-based-military-plotting. The military named entity, as the main carrier of the information, should be the key to military information extraction. So named entity recognition (NER) plays a significant role in these applications. In this paper, a process that was composed of CRF model and Rule-based method was proposed. With external lexicon feature added and rules applied the efficiency of named entity recognition is highly improved. Experiments show that great optimization was achieved.