Bilingual Structure Alignment Based on Monolingual Parsing

Aligned bilingual corpus has been recognized as a valuable resource for knowledge acquisition in machine translation and many other NLP tasks In this paper, a new method for automatic structure alignment of bilingual corpus is proposed The method is based on a bilingual language model-inversion transduction grammar By integrating a parsing constraint function into the bilingual language model, bilingual structure alignment is implemented using only one language parsing and word alignment results Compared with the traditional "parse parse match" procedure, the monolingual parsing supervised method is particularly useful when there is no robust parser for one of the languages The preliminary experiments show that the proposed method preserves the grammaticality of structure alignment and achieves good performance in accuracy Although this paper is related to the structure alignment of English and Chinese corpus, the proposed method is also applicable to other language pairs