Abstract: Machine translation involves computerized translation between natural languages, which is a new research field of combining computer technology and linguistics Translation selection is a key problem and remains an open one in this field Based on linguistics and fuzzy mathematics, a fuzzy classification model is constructed by evaluating the feasibility of target word combination as its kernel The translation selection experiment for an English Chinese machine translation system proves the satisfactory performance of this model The model simultaneously disambiguates all the words in the sentence scope, and thus is a proper solution for large scale machine translation system development