two novel model architectures for computing continuous vector representations of words from very large data sets计算词的连续向量表示的两种新模型架构large improvements in accuracy at much lower computational cost算力需求降低，