Little attention has been paid

in previous architecture design of computer assisted vocabulary learning (CAVL) systems

to the semantic connectedness among target words and its possible effect on learning. However

vocabulary learning researches indicate that different strategies should be adopted to process semantically related words at different stages of vocabulary learning. To improve current CAVL systems

this paper formulated and designed a scientific and intelligent learning scheme for CAVL system through word-embedding based word semantic similarity computation and lexical semantic network construction

as well as the tracking of learner usage behaviors. The study demonstrates that word embedding technology has great potential in application to CAVL.