Vocabulary recommendation for primary school compositions using natural language processing

For the study, corpora were built from student compositions, Chinese texts, and essential vocabulary lists. Data structures for individual words were created, and relationships between words were mapped using a Neo4j knowledge graph. Example sentences were generated by Alibaba Cloud’s Qwen, ensuring suitability for elementary students. At the same time, a survey was conducted among 100 teachers and students to evaluate the system’s effectiveness. Results indicated significant user satisfaction, with 75.9% finding it helpful. Indeed, there are still many shortcomings in this vocabulary system, such as: clearer polysemous word explanations, personalized recommendations, and enriched vocabulary resources using AIGC technology, etc.