| Popis: |
This study aims to investigate ChatGPT's ability to comprehend input from non-native speakers, specifically those learning English as a second language, with Japanese speakers serving as the model population. The experiment examines how ChatGPT evaluates the difficulty levels of the Test of English for International Communication (TOEIC), which is widely taken by non-native English speakers of varying proficiency levels. This study also measures ChatGPA’s response to discourses produced by non-native speakers - one containing grammatical errors and the other incorporating sociolinguistic or strategic competence expressions. The findings indicate that ChatGPT demonstrates proficiency in lexical and grammatical features and can comprehend non-perfect grammatical discourses produced by non-native speakers. However, ChatGPT does not accurately perceive the varying degrees of difficulty experienced by non-native speakers. Moreover, while ChatGPT can comprehend sociolinguistic expressions when the context is clear, its understanding of other communication strategies employed by non-native speakers is relatively limited. |
