Application of Large Language Models in Foreign Language Vocabulary Teaching

Traditional foreign language vocabulary teaching struggles with static materials and decontextualized learning, limiting learners' practical application. This study evaluates GPT4 and DeepSeek in teaching English near-synonyms (“fare” vs. “fee” and “promise” vs. “assure”). Through the qualitative and quantitative analysis of the generated exercises, results reveal GPT-4's strength in semantic expansion and contextual adaptability, while DeepSeek reinforces grammatical accuracy through structured rules. A hybrid approach combining both models bridges foundational knowledge and pragmatic fluency, enabling learners to shift from memorization to contextual mastery. However, challenges include technological gaps, teacher training, and AI biases. This research highlights AI's role as a supplementary tool, advocating human-AI collaboration to enhance cross-cultural competence and effective learning.