A Study on the Synergistic Development of Declarative Lexical Knowledge Representation and Cognitive Abilities in Adult EFL Learners Empowered by Word Embedding Models in the Era of Data Intelligence

Abstract

In recent years, artificial intelligence technology has made significant strides in the field of education. Word embedding models, as a type of language generation model based on natural language processing, possess the ability to engage in real-time dialogue with students, offering a new form of support for vocabulary learning among English learners. Using the PyTorch deep learning framework, we constructed a word embedding model based on neural networks to design a word association memory system environment and conducted a 12-week experimental study involving 60 adult EFL learners. The results show that the average total scores for the participants’ declarative vocabulary knowledge breadth and depth tests were 92.11 and 105.19, respectively, and that the word embedding model was able to enhance the development of cognitive abilities. Furthermore, at the macro level, the scale of the second-language vocabulary semantic network expanded, its diameter increased, and its small-world characteristics were enhanced, while the network’s density and centrality decreased. At the macro level, the network’s modularity increased, and semantic partitioning became clearer. At the micro level, central words remained relatively stable, but the specific positions of most words within the network fluctuated significantly; English proficiency did not affect vocabulary connections. Strategic awareness can be strengthened, learning interest stimulated, language input increased, vocabulary strategies optimized, language output emphasized, and cooperative learning encouraged.