Learning Synchronous Pattern Grammar for Machine Translation

• Main Authors: Yen, Tzu Hsi, 顏孜羲
• Other Authors: Chang, Jason S.
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/09993580682099109343

碩士 === 國立清華大學 === 資訊工程學系 === 103 === In this paper, we introduce a method for learning Synchronous Pattern Grammar (SPG) for assisting learners in translating sentences from one language into another. In our approach, we learn English pattern grammar from a given corpus, and then project the pattern grammar to Chinese through a parallel corpus with alignment annotation. The method involves converting English sentences into sequences of phrase chunks, converting phrase chunks to pattern elements, and extracting salient patterns for content words (verbs, noun, and adjective). The method also involves developing a machine-translation based Chinese word segmenter, developing a base phrase clunker, and converting bilingual phrases to synchronous grammar patterns. With synchronous grammar patterns, we present an interactive writing environ- ment, WriteAhead for translators, that automatically extracts and displays relevant synchronous grammar patterns with examples to prompt the user as they translate or mouse around a translation draft during editing.