ỨNG DỤNG DEEPSEEK NHƯ CÔNG CỤ HỖ TRỢ PHÁT TRIỂN KỸ NĂNG NÓI CHO SINH VIÊN CHUYÊN NGÀNH NGÔN NGỮ ANH TẠI TRƯỜNG ĐẠI HỌC THỦ ĐÔ HÀ NỘI

Abstract

This study investigates the effectiveness of DeepSeek, a large language model, as an intelligent tutoring system to enhance speaking skills among English Language majors at Hanoi Metropolitan University. Addressing persistent challenges in oral proficiency development, particularly limited personalized practice opportunities and immediate feedback in traditional classroom settings, the research employs a mixed-methods approach combining quasi-experimental design with qualitative data collection. Eight second-year students were divided into experimental and control groups for an eight-week intervention. The experimental group engaged in structured speaking practice with DeepSeek, utilizing its interactive dialogue, pronunciation analysis, and impromptu topic simulation features. Pre-tests and post-tests using standardized IELTS Speaking rubrics measured progress in fluency, accuracy, pronunciation, and coherence. Supplementary qualitative data from semi-structured interviews and learner journals provided insights into students' perceptions, motivation, and challenges. Results demonstrated that the experimental group achieved statistically significant improvements in overall speaking competence (average score increase of 13.1%) compared to the control group (4.5%). Qualitative findings revealed DeepSeek's role as a low-anxiety practice environment, fostering learner autonomy and providing readily accessible feedback. The study concludes that strategic integration of DeepSeek as a supplementary tool can effectively complement formal instruction and accelerate speaking skills development for English majors.