초록 열기/닫기 버튼
Kim, Dong-Sung. (2016). English Caption Writing Assessment Using Abstract Meaning Representation. The Linguistic Association of Korea Journal, 24(4), 235-260. Since story-telling has been used in evaluating the development of language skills, English language proficiency test such as TOEIC includes a caption writing test. This paper investigates how linguistically motivated features are used for automatically scoring a picture-description writing test. Specifically, we design to build scoring models with features under the principles of relevancy, appropriateness, and task-detailed description. For the experiment, we gather the caption writing corpus upon several images. We statistically compare different performances among 9 statistical assessment factors, revealing that Abstract Meaning Representation (AMR) produces the best results in predicting human raters’ scores. AMR shows the best performance in capturing the similar logico-semantic structure(s) among various sentential forms.
