Multilayer Anchor Alignment in AC-E Parallel Corpora of Chinese Tea Classics

Chinese tea literature has not made its appearance in the existing corpora. Bilingual corpus of ancient Chinese and English (AC-E) also wait to be extended for purposes such as CAT and designing educational software for Confucius Institutes all over the world. This paper aims at such a corpus by demonstrating multilayer anchor-points in improving the alignment accuracy in a bilingual parallel corpus of tea classics. An experiment is carried out with four layers of “anchor points”. Technical terms as the first layer are extracted with Term List of Winalign module in Trados. The second is register-specific words with 1:1 co-occurrence frequency in SL and TL. The third and fourth are composed respectively of proper nouns and transliterated Chinese-unique words. Statistics show that the alignment accuracy keeps increasing with the step-up of each layer. Since such anchor-points are typical in ancient Chinese classics, this method can be generalized in relevant fields.