A STUDY ON THE RECOGNITION OF THE SYNONYMOUS AFFIRMATIVE AND NEGATIVE CONSTRUCTIONS

Abstract

In this paper, we referred to such expressions like "cha dian'er +X" as "synonymous affirmative and negative constructions". As the usage of the constructions were completely complicated, it might make it difficult for machine to understand such constructions, which required us to study how to recognize them. Firstly, we collected and analyzed the synonymous affirmative and negative constructions existing in modern Chinese language based on the corpus and researches of the predecessors. Then by annotating and analyzing the corpus which involved 9 construction marks such as "cha dian'er(差点儿)" selected from the overall collections, this paper tried to illustrate syntactic rules for each constructions. Furthermore, we carried out 3 pairs of experiments based on above annotated corpus, to test the recognition ability of RoBERTa+Fine-Tuning model, and to discover whether there were same recognition knowledge between different constructions.