Finding translation correspondences from parallel parsed corpus for example-based translation

摘要

This paper describes a system for finding translation correspondences from parsed parallel corpus that is the paired dependency structures of a source sentence and its translation in a target language. At first, the system we have developed finds correspondences by consulting a source-to-target word dictionary, and then finds the other remaining correspondences based on dependency structure and the balance of all correspondences. The system achieved about 70 coverage and 80 accuracy. For finding correspondences, the fact that the source sentences in parallel corpus correspond suitably with the target sentences, is available. This method can collect more correspondences than a statistical approach.