英文摘要 |
In this paper, we report our experiment on a modular statistical approach to machine translation system. The experimental MT system consists of modules implemented by statistical methods to handle different level of linguistic analysis. The overall architecture of the system resembles that of a transfer-based MT system, but with less explicit expert knowledge involved. Five hundred simple bilingual sentences with main verbs restricted to 30 commonly used verbs are used as training data. These sentences are syntactically and semantically tagged to provide statistical data for case role analysis and transfer. A bilingual dictionary and collocation data from a corpus of Chinese news are used in target generation. The system is tested against the original 500 sentences and additional 100 sentences with promising results. |