摘要
一种基于语料库的军用文书自动生成中表层实现的概率模型。运用最大熵模型,把N元文法语法模型获取的生成文本中词汇间的同现概率信息、句法依赖树中的句法信息和属性信息进行有机结合,生成文书文本。实验结果证明,该模型可以逐个词汇地生成满足要求的文本。
A probability model based on the corpora for a surface realization in automatic generation of military document. This model organically combines the coincidence probability information between the lexis in the generation text, which was obtained from the N-gram by the maximum entropy model, with the information from the syntax and features in the dependence tree. Experiment result shows that the model the field of vocabulary may be produced the text to meet requirement one by one.
出处
《兵工自动化》
2010年第11期60-62,73,共4页
Ordnance Industry Automation
关键词
军用文书
表层实现
N元文法
最大熵模型
military document
surface realization
N-gram
maximum entropy model