Japanese-Chinese Machine Translation of Japanese Determiners Based on Templates

Abstract

The machine translation of Japanese sentences with determiners, like “shika...nai”, “tyoutto...dakedeha”, “tada...dake” and so on, are more special and regular on sentences structure. The research collects and classifies the Japanese sentences which contain the determiners. The classification is carried out by according to the characteristics of Japanese sentences and translation habit of Chinese sentences. Through further abstraction and simplification, translation templates are extracted by gathering grammar rules information, studying syntax and analysis the collocation mode of sentences. Those determiners express confirmed meaning, and the corresponding translation Chinese sentences have the same characteristic. By analyzing the sentence characteristics with determiners and formalizing the sentences structure, the translation templates are abstracted. By investigating the structure characteristic of original sentences with translation templates, the similarity algorithm was defined. The threshold value of the similarity calculation was obtained by preliminary experiments, and the experiments of Japanese-Chinese translation are carried out by a small corpus. The experimental results for several kinds of Japanese sentences with determiners show the translation accuracy rate is 68.6%, template coverage rate reach 83.3%. At last, through the analysis for the translation errors, following conclusion is drawn: the results of morphological analysis are erroneous, because the error of word segmentation the part of speech tagging also are erroneous, result in the grammar structure cannot match with templates; the original sentences are long and especially complex sentences; the templates are too complicated; the similarity calculation method needs to discuss further, and so on.