To improve the accuracy of the translation of the ChineseMongolian language machine translation system, this paper proposed the methods that combined the phrasebased statistical machine translation and template based machine translation to automatically extract template structure. Template matching was the first step in decoding and the template structure was applied for translation, and then according to Beam Search algorithm a follow-up translation was carried out. Thus, the method can effectively resolve Word Order errors in a single statistical translation. Take a Chinese-Mongolian translation as an example, the experimental results show that this method can effectively improve translation effects. In the area of agricultural language Translation, agriculture phrase templates were added, and the translation efficiency increased by 10% than Och‘s phrase-based statistical machine translation methods.