在传统的自适应研究方法中，数据自适应的方法通过设计相似度函数，在训练数

在传统的自适应研究方法中，数据自适应的方法通过设计相似度函数，在训练数据中选择和待翻译测试集文本领域更为一致的源领域数据，这种领域更为“相似”的数据选择方法不仅可以缩小训练规模，也能有效的提高翻译质量。子模型自适应的方法将训练数据根据不同的领域特征划分为几类，针对原训练集的不同子集分别训练子翻译模型，然后为每个子翻译模型调整权重，使翻译系统针对不同领域的待翻译文本仍然有较好的翻译表现。

In the traditional adaptive research methods, the data adaptive method selects the source domain data that is more consistent with the text field of the test set to be translated in the training data by designing the similarity function. This more "similar" data selection method can not only reduce the training scale, but also effectively improve the translation quality. The sub model adaptive method divides the training data into several categories according to the characteristics of different fields, trains the sub translation models for different subsets of the original training set, and then adjusts the weight for each sub translation model, so that the translation system still has a better translation performance for the text to be translated in different fields.

In the traditional adaptive research methods, the data adaptive method selects the source domain data which is more consistent with the text domain of the test set to be translated from the training data by designing the similarity function. This data selection method with more similar domain can not only reduce the training scale, but also effectively improve the translation quality. Sub-model adaptive method divides the training data into several categories according to different domain characteristics, trains sub-translation models for different subsets of the original training set respectively, and then adjusts the weight for each sub-translation model, so that the translation system can still have good translation performance for texts to be translated in different domains.