Machine Translation of Chinese–Hindi Simple Sentences Using Moses

In this paper, a few simple Chinese sentences have been translated into Hindi using Moses, a statistical machine translation system. A total of 3015 sentence pairs of Chinese–Hindi are considered for training the model. For training the model, data are preprocessed which includes the following steps: tokenization, truecasing and cleaning of corpus. Since Chinese language does not have space delimiters between words, so before tokenization, it requires segmentation and chunking. For this, Moses does not provide a tool. A separate tool for segmentation of Chinese text is used. In our case, we have only considered simple sentences limiting the length to 20 words; thus, we do not require cleaning of the corpus. The language model is generated using 4-gram technique. Testing of 35 Chinese–Hindi simple sentence pairs shows an accuracy of 67.57%. Also, it gives a BLEU score of 0.5 which indicates good translation quality, but needs further improvement.