本文从介绍转换网络和递归转换网络开始,给出了由Woods首先提出,最后由斯坦福大学Winograd教授完整描述的扩充转换网络ATN方法。这是研究自然语言理解的著名方法。文中还仔细讨论了寄存器结构、词类、字典及不同种类弧的分析方法,最后列出了一个较完整的英语ATN文法,对于我们进一步研究汉语的自然语言理解有参考价值。 This paper begins with introducing the transformation network and the recursive transformation network, and presents the ATN method which is first proposed by Woods and finally described by Professor Winograd of Stanford University. This is a well-known method of studying natural language understanding. The article also carefully discusses the register structure, parts of speech, dictionaries and analysis methods of different kinds of arcs. Finally, a complete English ATN grammar is listed, which is of reference value for us to further study Chinese natural language understanding.