Rhetorical Signs as Clues for Improving the Accuracy of MT - A Case Study from Classical Chinese to English

The paper intends to address a problem in classical Chinese-English machine translation, which still waits anxiously for joint efforts from linguists and computational engineers to get developed. Google online translation system and SystranBox are tried out on decrypting The Classics of Tea, and both versions reveal problems potentially typical in this field. One obvious defect is their inadequate capacity in recognizing rhetorical devices such as “repeated” words in parallelism and antithesis with concrete or redundant information. It is expected that such devices could be used as indicative clues for bilingual alignment and generative algorithm to increase the effect of classical Chinese-English machine translation.