| The thesis incorporates two main topics:a preliminary evaluation of online machine translation systems and a concrete evaluation experimented on a sample test case. Both are quantitatively evaluations, with the former referencing the model provided by John Lehrberger in A Synthesis of Evaluations of Machine Translation Systems (1981) and the latter taking the formula in Target Text Quality Requirements for Translation Service (2005). Three online machine translation systems are evaluated:Worldlingo, Google Translate and Youdao Translate.Four questions are discussed in the thesis. Can online machine translation systems be used in localization? What is the percentage of sentences with the passing mark? For sentences with passing mark, what is the gross error rate? If the results of online machine translation systems can be used in localization, how it can be used?100 sentences from product manuals are randomly selected to make the evaluation and they are from 5 fields:domestic appliance, pharmaceutical, machinery, software and chemical engineering, with each 20 sentences. A preliminary assessment is firstly taken to differentiate valid and invalid sentences. Two criteria are inspected in the assessment, which are fidelity and intelligibility. They are divided into 5 levels, with 3 the passing mark. If the average values of the 100 sentences exceed the passing mark in both fidelity and intelligibility, then the online machine translation can be used in localization. Then, the percentage of sentences with passing mark is calculated. Next, a concrete evaluation is taken to make out the gross error rate of sentences with passing mark. At last, the application of online machine translation systems in localization is discussed.According to the evaluation, Google Translate can be used in localization, whose average values of the 100 sentences are respectively 3.49 and 3.2 in fidelity and intelligibility. Worldlingo and Youdao Translate fail with the values of 2.17 and 2.42, and 3.29 and 2.71 separately. For the percentage of sentences with the passing mark, Worldlingo is 23%, Google Translate accounts for 64% and Youdao Translate holds 49%. For sentences with passing mark, the gross error rates of Worldlingo, Google Translate and Youdao Translate are 1.83%.2.24% and 1.64% in turn.The findings of the thesis can be used to the further research of online machine translation systems, and it has some significance for the use of online machine translation in localization. |
PDF Full Text Request