语义省略“的”字结构中心语自动补全研究

Research on Automatic Head Completion of “de” Construction with Semantic Ellipsis

省略作为一种复杂的语言现象,一直是汉语语义研究的重要课题。如何将省略的语义内容补全以还原句子的完整语义也是机器理解自然语言的关键。本文基于AMR(Abstract Meaning Representation)语料统计得到含有语义省略信息的"的"字结构,针对特指性和泛指性省略类型,采用特殊句式核心谓词对应主宾语和基于谓词框架抽取核心谓词搭配必有论元成分的策略,确定补全的语义类别。研究结果表明,该方法对于省略"的"字结构中心语补全有较好的效果,可为自动理解深层语义奠定基础。

Ellipsis as a complex language phenomenon has always been an important topic in Chinese semantic research. How to complete the omitted semantic content to restore the complete semantics of the sentence is also the key to the machine understanding natural language. Based on the statistics of AMR(Abstract Meaning Representation) corpus, this paper obtains the "de" construction with semantic ellipsis information and the model uses subject or object corresponding to the core predicate to determine the conceptual category of the ellipsis head for the "de" construction of the special sentence pattern. We extract the candidates with argument components from the core predicate of the "de" construction by using the predicate frame to determine the semantic concept type of the ellipsis. Experiment shows that this method has a good effect on head completion of "de" construction with semantic ellipsis and lays the foundation for automatic understanding of deep semantics.