A Study on Machine Translation of Register-Specific Terms in Tea Classics

The rich heritage of Chinese tea culture has attracted an increasing number of people in the world, but the translating of such classical and specialized literature proves to be extremely arduous. Machine translation (MT) is introduced to facilitate the decrypting process. However, when the popular online translation Systran is tried bi-directionally on a high-frequency wordlist from 24 ancient tea documents, the version reveals some defects, especially the recognition of register-specific terms. To address this elementary yet essential problem, a parallel translation corpus is constructed. Statistic method of automatic extraction is used to produce a bilingual inventory of equivalent term pairs. To further improve the efficiency of MT, a descriptive term database is built on the basis of the previous work. It is expected that, with contextual clues taken into the decrypting process, later machine-aided translation of documents in the special sphere could become more standardized, consistent and systematic.