基于实例的日中机译系统的模型构造

对于基于实例的机器翻译，最困难的是如何从多个翻译实例中选出最相信的样例，本文对日中机器翻译系统提出了一种基于概念分类与附属词的方法来计算相似性，该方法充分体现的名中词间的组合搭配关系，可适当缩减实例库规模，简化权值的计算，取得了较好的效果。

Japanese-Chinese machine translation module's structure based on examples

For the machine translation based on examples, the most difficult problem is how to select a most similar example from multiple examples. This paper presents a method based on conception classification and relative word to calculate the similarity of the example. This method reveals the relation of compositing and collocation between the words in the sentence. The examples' base and the weighting calculation can be reduced appropriately and the satisfactory result is obtained.