Semantic analysis of chinese prepositional phrases for patent machine translation

Abstract

In Chinese patent texts, prepositional phrases(PP) are quite long with complicated structures. The correct identification of PP is very important for sentences parsing and reordering in machine translation. However, existing statistical and rule-based methods perform poorly in identifying these phrases because of their unobvious boundaries and special structures. Therefore, we present a method based on semantic analysis. Chinese prepositions are divided into two categories due to their semantic functions, and more contextual features are employed to identify the phrase boundaries and syntax levels. After integrating into a patent MT system, our method has effectively improved the parsing result of source language. © Springer-Verlag2013.