A new Chinese-English machine translation method based on rule for claims sentence of Chinese patent
Wen Xiong, Yaohong Jin · 2011
Machine translation (MT) is a hot research in artificial intelligence (AI) and natural language processing (NLP), which has mainly four directions such as statistics-based, instance-based, rule-based, and interlingua-based. To improve the quality of translation for Chinese patent, the paper proposes a new MT method based on rule for the claim sentences of Chinese patent. First, it analyzes the composition of the claims, and presents a Backus-Normal-Form (BNF) expression to parse them into fixed and variable parts. Then, it translates the fixed with prepared texts and the variable using a MT system based on syntax-rule. Finally, it integrates them into a full sentence. Experiments show that the BNF expression has a coverage ratio of 94.74% for claims on a Chinese patent dataset collected manually, which indicates its availability.