On January 1, the 57th ACL Branch, the First International Workshop on Designing Meaning Representations (DMR), was successfully held in Florence, Italy. The ellipsis in Chinese AMR corpus, a collaborative paper by Liu Yihuan and Yan Peiyi, graduate students of our institute, was recruited and Liu Yihuan attended the Italian Conference.Poster display was performed.
This workshop aims to bring together producers and consumers of semantic representation to discuss and understand the key issues of semantic representation through their interaction.The workshop hopes to promote the critical evaluation of existing semantic representations by semantic representation researchers, design a new generation of semantic representations, and explore the opportunities and challenges of designing and using semantic representations in multilingual environments.The workshop was chaired by Xue Nianwen, an associate professor of computer science at Brandeis University, USA. A total of 22 papers were included.
Liu Yihuan et al. based on a new semantic representation method, abstract semantic representation (AMR), this paper studies ellipsis in Chinese sentences.According to the linguistic expression of ellipsis in Chinese, ellipsis can be divided into three categories: ellipsis of antecedents with consistent semantics, ellipsis without antecedents, and ellipsis of antecedents with inconsistent semantics.According to different ellipsis phenomena, AMR uses different methods to restore ellipsis in sentences.On the basis of 5000 Chinese AMR tagged corpus, this paper statistics the distribution of ellipsis in Chinese. 54% of Chinese sentences have ellipsis. Among them, 92% have antecedents with identical semantics. It is found that ellipsis sentences are about 6 units longer than ordinary sentences (words or concepts).In addition, the paper also explores the nature of ellipsis and finds that ellipsis is mostly a complete phrase, but sometimes it may be a part of a phrase or a headword.
During the exhibition, many participants expressed their interest and affirmation in the work, discussed the treatment and distribution of ellipsis in Chinese and other languages, and also discussed some issues related to AMR, such as predicate in Chinese PropBank.The one-hour exhibition not only made the participants fully understand our work, but also benefited us a lot. Our computational linguistics will continue to research and explore Chinese AMR.