IHSMT中的实例优化策略

THE STRATEGY OF CASE OPTIMIZATION IN IHSMT

在基于实例的机器翻译方法中 ,通常采用双语句子实例的形式 ,但由于自然语言表达的无限多样性 ,使得这种存储粒度过大的句子级翻译实例的利用率较低 ;而在基于规则的机器翻译方法中 ,规则和词典具有一定的抽象性 ,其重复利用率高 ,但其中存在大量的歧义 .提出了一种 IHSMT中的实例自动优化算法 ,该算法引入了实例粒度的概念 ,根据规则和实例使用率 ,将粒度大的实例分解为较小的单元 ,并建立相应的词典 ,以提高实例的使用频率和检索效率 ;同时在翻译过程中 ,根据用户修改的统计和上下文信息 ,对粒度小并具有歧义的实例进行合并 ,以提高翻译质量和推理效率 ,从而使实例粒度更加合理 ,存储结构更优化 。

In case based machine translation, the bilingual case of sentence is always used. The utilization of sentence case with big granularity is low, because of unlimited multiplicity in natural language. In rule based machine translation, the knowledge of rule and dictionary is abstract, and their usage is frequent, but there is much ambiguity. Proposed in this paper is an automatic optimization algorithm of case in IHSMT. The algorithm defines the concept of case granularity and decomposes the big case granularity into small ones, and then the related dictionary is built, so the frequency of utility and efficiency of retrieval are improved. At the same time, the small granularity with ambiguity can be combined according to the modification of users and the information of context in translation, so the quality of translation and efficiency of reasoning are better. The case optimization makes the granularity and the storage of case more reasonable, and the efficiency of the system is improved.