Identification of comparative sentences with adjective predicates in patent Chinese-English machine translation

Comparison is a common expression to indicate the relations of different objects and it exists in almost every language. In Chinese patent texts, there are lots of comparative sentences, and more than 70% of the sentences whose predicates are adjectives. In particular, the structure like “*** (be less than)” is quite rare in the previous researches, which is a key and difficult point of our work. So the study on identification of these sentences in patent texts is very significant and can contribute to the patent Chinese-English machine translation. In this work, we applied the method based on semantic rules to analyze and identify comparative sentences with adjective predicates in patent texts. Various linguistic features of these sentences have been explored, such as comparative keywords and patterns, based on which a lot of semantic rules for identification were done. Experiment verifies good performance of our method compared with Google translator and Baidu translator. The method improves the performance of patent machine translation effectively.