Alignment algorithm for bilingual paragraphs

Abstract

The invention provides a method for carrying out syntagm alignment by stacking Bilingual Corpus, which is the basis of example-based machine translation (EBMT), provides an alignment module of bilingual syntagm of English and Chinese based on anchor word-pair and corresponding alignment algorithm and solves the data sparse problem of middle or small Corpus. The ambiguity in syntagm segmentation is delayed to be removed by the system at the time of syntagm alignment, thereby improving the correctness of syntagm segmentation.

Description

technical field [0001] The invention patent relates to English-Chinese bilingual understanding technology in natural language understanding. Especially segment alignment technology Background technique [0002] In recent years, with the development of corpus linguistics, the example-based machine translation (Example-based MT) method has become one of the new ideas of machine translation. The EBMT system stores a large number of segment-level aligned bilingual sentence pairs in advance, that is, a bilingual corpus. When translating, the system only conducts a shallow analysis of the translated sentence, divides it into segments, and then finds the best translation for each segment from the bilingual corpus according to the context, arranges them in a certain order, and finally generates translated sentence. This method avoids many difficulties in traditional translation methods (such as syntactic analysis, word meaning recognition, etc.), and has certain practicability, e...

AI Technical Summary

Problems solved by technology

This makes the probability data in the model (such as the probability of any English word translated into any Chinese word) have serious data sparsity problems

Images