Earticle

초록

영어

Event-related potential (ERP) has been widely utilized to detect cognitive behaviors. Recently, second language (L2) processing has been proven to detect brain responses similar to native (L1) from language-related ERP components such as N400/P600. In this study, we aim to detect syntactic violations from individual differences based on ERP components. To this end, we conduct experiments to discriminate detected ERP components using machine learning by proficiency levels. We recorded EEG signals from 22 participants, German L2 as Korean, while reading sentences, some of which included syntactically anomalous sentences. After data preprocessing, we evaluate the individual differences in discriminating the syntactic violation: ERP component analysis, classifying syntactic violations and comparing proficiency levels. As a result, we confirm the detection of the P600 component from advanced learners but not beginners. Performance evaluation shows that the Long-Short Term Memory (LSTM) based on a subject-independent approach has the highest accuracy of 68.85% for beginners and 69.99% for advanced learners, respectively. Additionally, we also achieve the best accuracy of 76.66% using the LSTM regarding the proficiency levels. In conclusion, we confirm that L2 learners are more similar to natives as their proficiency levels increase. Furthermore, we will closely investigate the brain responses between L1 and L2 to determine whether achieving the same brain response as a native when language comprehension is possible.

목차

Abstract
I. INTRODUCTION
II. MATERIALS
A. Institutional Review Board (IRB)
B. Subjects
C. Materials
D. Experimental procedure
E. EEG recording and preprocessing
III. METHODS
A. Classification models
B. Feature extraction
C. Experimental results
IV. EXPERIMENTAL RESULTS
A. Event-related potential analysis
B. Comparison of classification based on syntactic violation(correct vs. incorrect) with subject-independent
C. Comparison of classification based on proficiency level (Beginner vs. Advanced learners)
V. CONCLUSION
REFERENCES

저자

• Wonyoung Lee [ Vision AI Business Team LG CNS Seoul, South Korea ]
• Guiyoung Son [ Dept. Software Sejong University Seoul, South Korea ] Corresponding Author

참고문헌