When translating a file, the first thing we think of is that the translation must be very accurate. There are many file translation service providers on the market, including not only human translation services but also machine translation services, and in the last 2-3 years, AI translation services as well. Today, I will compare DeepL and Claude translations to see which one is more accurate.
The following is my process of comparing DeepL and Claude translations using ITransBook.
1.Open our ITransBook translation software.
2.Click the File Translation icon on the left.
3.Click the + sign in the upper left corner to create a new project.
Enter "Comparison of DEEPLE and Claude engines" for the project name. Select "English (American)" for the source language, choose "Target (1:N)" for the target option, select "Chinese (Simplified, 简体中文)" for the target language, and choose DeepL and Claude as the translation engines.
4.Click "API Settings" on the right.
I have already purchased a credit package, so I don't need to configure the API keys for DeepL and Claude. I can choose the default model, and here I will select "claude-sonnet-4-6".
Click the "Save" button.
5.Drag the file to be translated into the file translation area.
This is a screenshot after the translation is completed.
Click the "More" button, then click the "Export multilingual document" button to export the comparison translation.
Check the three displayed options and click the "Export" button.
Click the "OK" button to proceed to the comparison translation folder.
We double-click the comparison translation file to open it and begin comparing the DeepL and Claude translations.
Both DeepL and Claude translated "token" incorrectly. On March 25, the China National Committee for Terminology in Science and Technology officially issued an announcement, preferentially recommending "词元" as the standard Chinese term for "token" in the field of artificial intelligence. In translating "companies need to have both human capital and in-house AI capabilities...", Claude's translation "需要同时具备" is more specific and accurate than DeepL's translation "需要...". For this passage, I personally think Claude's translation is better.
DeepL translated "firms across the economy" as "各行业企业", which is easy to understand, and "workers" as "员工", which fits the context better, superior to Claude translating "workers" as "劳动力". For "a lot of displacement", both DeepL and Claude translated it poorly, with DeepL's translation being further from the original meaning; it should be translated as "大量岗位变动" instead. For this passage, I personally think DeepL's translation is better.
For the translation of the word "narratives", DeepL's translation as "说辞" is better, carrying an emotional tone, but for the subsequent "just narrative", DeepL translated it as "叙事", showing an inconsistency in terminology translation. For these two passages, I personally think DeepL's translation is better.
For "reduce prices for customers", DeepL translated it as "为客户降低成本", while Claude translated it as "为客户降低价格". DeepL's translation is more accurate and flexible; the passage revolves around costs, so it should be reducing costs. For this passage, I personally think DeepL's translation is better.
Overall, DeepL was more accurate in translating this interview article, while Claude still has room for improvement in terms of word choice.
Top comments (0)