Research And Implement Of Statistical Machine Translation System On Handheld Device

With the development of economic globalization, international exchanges and cooperation have become increasingly frequent and in-depth. The language differences have been a serious obstacle. Therefore, the development of high-quality and practical machine translation system becomes important. And high efficiency and high-quality machine translation system on handheld devices becomes more practical value.At present, the machine translation system on handheld device has been widely used, but most of the approach is grammar rule-based. Phrase-based statistical machine translation system on handheld devices is still in its infancy, especially the Mongolian machine translation.In this paper, we focus on the characteristics of handheld devices, the design part of the machine translation system decoder. We design a decoder for handheld devices, using a fast decoding method-beam search decoding. We use the open-source statistical machine translation software to train the language model and translation model, and the decoder to decode. In addition, we optimize the phrase table from the content and structure.The experimental results show that the decoding quality of the translation is acceptable. And using the optimized phrase table, we can effectively improve the speed of translation.