The purpose of this study is to present a diagnostic evaluation tool that can more accurately determine the academic performance level in Korean language proficiency of foreign undergraduate students studying at domestic universities. The diagnostic evaluation diagnoses the academic Korean language proficiency of foreign undergraduates, and provides the diagnosis results to foreign undergraduates, professors, and university institutions so that teaching activities can be conducted efficiently. In this study, it was considered that the ‘Basic Academic Ability Diagnostic Test for the Korean Language Subject' was suitable as a test tool for diagnosing to diagnose the academic achievment of foreign undergraduates. This diagnostic test is a diagnostic tool designed for Korean students in Korea, and it not only continuously verifies its reliability and validity, but also includes background knowledge that Korean university students in Korea are basically acquiring. To prove that the 'Basic Academic Ability Diagnostic Test' is suitable as a diagnostic evaluation tool for foreign undergraduates, the correlation with the Korean language achievement evaluation was analyzed. The correlation between diagnostic evaluation and other items showed a significantly positive correlation. When the diagnostic evaluation score was good, the pre-test and post-test scores of Korean language proficiency, and the TOPIC level was also high. This result proves that the basic academic ability diagnostic test tool is very suitable as a diagnostic evaluation tool for foreign undergraduate students. Also, the analysis result of the pre- and post-tests and the TOPIK test showed a very significant and strong positive correlation with the post-test, and the higher the TOPIK grade, the higher the score of the pre‐test.
본 논문에서는 한국어 쓰기 교육을 위한 교육용 게임으로서 text-to-scene 모델을 제안한다. 사용자 입력 문장에 대해 즉각적인 피드백을 제공하기 위해서, 제안하는 text-to-scene 모델은 사용자 입력 문장의 의미구조를 분석하여 장면을 자동으로 생성한다. 사용자가 문장을 교정하면서 자연스럽게 한국어 쓰기 능력이 향상될 수 있도록, 제안하는 text-to-scene 모델은 정답 장면과 사용자 입력 문장에 대한 장면을 함께 제공하므로, 사용자는 입력 문장의 어느 부분이 잘못되었는지 쉽게 파악하고 문장을 수정할 수 있다. 기존의 text-to-scene 연구에서 정교하게 장면을 생성하기 위해서는 필요한 자원 구축이 어렵고 처리 과정이 매우 복잡하다는 점을 고려하여, 제안하는 text-to-scene 모델은 외국인이 쉽게 이해할 수 있는 수준으로 장면을 단순하게 표현한다. 실험결과 격조사나 문장 요소 생략 등을 고려하여 15개의 단어를 임의로 조합하여 5개 어절 이내의 문장 1,048,576개 문장을 만들어 실험한 결과 제안하는 모델은 한 문장당 평균 14.94ms가 소요하면서 안정적으로 장면을 생성하였다. 외국인을 대상으로 한 작문 시험 결과, 종이를 이용한 작문 결과에서는 10개 문항 중 오답 문장이 평균 4.7개 나왔는데, 제안하는 text-to-scene 모델을 이용한 작문 결과에서는 한 문장 당 평균 1.56회의 미리보기를 하고 문장을 수정하여 0.87개가 나왔다.
A large number of foreign students enroll at Korean universities each year. Although a good command of the Korean language is crucial for them to be successful in their studies, and both university authorities and teaching staff now agree that students’ language ability needs to be tested before their admission, a valid and reliable test of academic Korean is still to be developed. The aim of this paper was twofold: to determine the Korean language proficiency level and language task types that are necessary to succeed at the university level in Korea; and to determine test specifications that could be used by examiners to appraise university candidates’ language skills. To develop such specifications, Common European Framework (CEF) descriptors of language proficiency were used. These descriptors can be used as indicators showing which language tasks a candidate should be able to do at a certain level. To create test specifications, academically relevant descriptors were selected and task types associated with each were analyzed. Many descriptors associated with the B2 level of the CEF were used to determine the language achievement level necessary for university students in Korea. Information about linguistic input and output associated with each language task type was also gained and applied from CEF descriptors.