为提升管制员培训效果,减少人物力成本,利用深度学习序列到序列框架,对陆空通话标准用语(英文)进行处理与建模,实现输入管制员文本指令,即可输出飞行员文本应答。首先模拟管制员飞行员对话用语习惯,创建航行进离场阶段英文数据集;其次建立陆空通话模型,并对模型进行优化和训练;最后通过相关指标评估模型效果。实验结果显示,模型应答具有较高准确率,证明论文方法在陆空通话领域具备实用性和有效性。In order to improve the effect of controller training and reduce the cost of human resources,the standard terms of ra⁃diotelephony communications are processed and modeled based on deep learning by using sequence to sequence framework.The model can output pilot text response by entering controller text instructions.Firstly,the English data set of the departure and arrival phase is created with the conversation habits of the controllers and pilots.Secondly,the radiotelephony communication model is es⁃tablished,optimized and trained.Finally,the effect of the model is evaluated by relevant indicators.The experimental results show that the model has a high accuracy in response,which proves that the method is practical and effective in radiotelephony communication.
机构地区中国民航大学空中交通管理学院
出处《计算机与数字工程》 2021年第3期466-470,共5页Computer & Digital Engineering
基金国家自然科学基金项目(编号:71801215)资助。
关键词自然语言处理 深度学习 陆空通话 空管模拟机NLP deep learning radiotelephony communication ATC training simulator
分类号O141.4 [理学—基础数学]
