This paper aiming at the problems of scattered operation types, poor extraction effect of learning resources and incomplete assignment of learning tasks in the online language learning platform designs a language assisted learning model under the background of artificial intelligence, proposes an English learning resource extraction algorithm based on the integration of LSA and N-gram models and then proposes a weighted multi-task learning model and it is optimized to improve the sparsity of learning variables. Finally, based on the construction of the assisted learning model, the online language learning system is designed, including demand analysis, overall design and architecture design, to improve the pertinence and efficiency of English learning.