Machine Translation Based On Case Frame

Adding semantic knowledge to machine translation has become an active area, many researchers try to add semantic information to the system of statistical machine translation, and they think semantic information is useful to machine translation. However, the construction of semantic information needs a lot of human resource and material resource, meanwhile Chinese semantic corpus is scarce, so the automatic learning method of getting Chinese and english semantic corpus is necessary.In this context,through integrating VerbNet,WordNet with FrameNet, enriching the FrameNet semantic information. At the same time, developing a system of semantic role labeling based on FrameNet to extract English-Chinese bilingual case. Then using the Bi-Case frame to design a system of machine translation.Semantic role contains the shallow semantic information, so it has been used in question answering system,information extraction,machine translation and other fields. Here using FrameNet to develop a system of semantic role labeling based on maximum entropy. The system accuracy rate is 77%,recall rate is 73.5% and F1 is 75.3%.Extracting parallel lexical units from HowNet and FrameNet. At the same time, using the system of SRL to labeling the English corpus of the bilingual parallel corpus, then labeling the Chinese corpus with parallel information, finally extracting Bi-Case Frame based on the bilingual parallel corpus with role.Eventually, this paper depicted a system of machine translation based on the Bi-Case Frame. The experiment showed that the system could improve the quality of translation when adding the Bi-Case Frame to it.