| 초록 |
Kang, Arum & Lee, Yong-hun. (2023). A study of discourse anaphora and definite NP in Korean: Utilizing deep learning models and neural machine translations. The Linguistic Association of Korea Journal, 31(4), 1-34. In this preliminary study, we investigate the phenomena of discourse anaphora and definite descriptions within the framework of the so-called donkey sentence. Unlike English, Korean allows for the expression of donkey anaphora using either the pronoun kukes it or definite noun phrases (bare NP or ku+NP). Employing neural machine translations and deep learning models, we examine the appropriateness of these two types of donkey sentences in Korean through the following procedure: Firstly, utilizing ChatGPT, we generate 60 sentences with donkey structures containing both pronouns and definite noun phrases. Secondly, we employ Google Translation and Papago to translate these sentences. Thirdly, we use KR-BERT to evaluate the acceptability of the translations. Finally, we conduct a statistical analysis based on the obtained acceptability scores. The results reveal that definite noun phrases are a more natural expression than pronouns in Korean donkey sentences. This novel finding suggests that the E-type approach would provide a better theoretical account than DRT (Discourse Representation Theory). |
