How to Use the DeepL Translation System
DeepL is an AI startup based in Germany that focuses on providing translation services using artificial intelligence. DeepL Translator was launched in August 2017 by DeepL GmbH, a startup supported by Linguee, and it supports translations between Chinese, English, German, French, Japanese, Spanish, Italian, Dutch, and Polish, utilizing neural networks to train its translation capabilities.
According to TechCrunch from the USA, tech giants like Google, Microsoft, and Facebook have applied machine learning to the translation field, but a small company named DeepL has surpassed them and raised the standards in this area. Its translation speed can rival that of larger competitors, but it is more accurate and detailed than any translation tool they have used.
Platform website: https://www.deepl.com/home
Official download address: https://www.deepl.com/app/
The platform interface is as follows:
The platform includes unlimited text translation, document translation, Linguee human translation database, and some other extended features.
Text Translation:
Document Translation:
Linguee Online Dictionary:
PC-side Translator Operation Interface (Free Version):
It should be noted that DeepL Pro is a paid service, with prices ranging from 6 euros to 40 euros per month. The cheapest personal annual plan is 6 euros, approximately 46 yuan, which is quite expensive. DeepL Pro provides users with unlimited text translation on the web version and application. The free translator has limitations on character count and translation volume, but DeepL also offers free translation, allowing users to directly use the free web translator, which should be sufficient for most users.
First, we need to download and install the DeepL translation software from the official website, then perform simple settings to enable direct access to DeepL translation software on the computer. Alternatively, we can copy the URL https://www.deepl.com/translator to open this online tool platform. DeepL is very intuitive to use; just open this webpage and paste the content to be translated. DeepL can translate online or use the MAC and Windows versions, supporting both text box and document translation methods.
2.1 Comparison of DeepL with Other Systems’ Translation Results
Below is an example from an article by the WHO, with April 4 being World Health Day, and the article from the World Health Organization is quite rigorous. Regarding this translation, there is a long sentence that has some differences:
Google translates it as “From history to today, nurses have been at the forefront of fighting epidemics and pandemics that threaten global health.” Bing translates it as “Historically, and today, nurses are at the forefront of combating epidemics and pandemics that threaten global health.” Google and Bing’s translations have a strong machine translation flavor, while DeepL translates it as “Whether in history or today, nurses stand at the forefront of combating epidemics and pandemics that threaten global health.” It is clear that DeepL achieves a fluent translation with accurate meaning. This “more natural” feel is a significant advantage of DeepL, which is also an easily perceivable difference for users.
What needs can current translators meet? We cannot expect the translations to be very precise; they should at least be fluent and logically correct, making it easier to refine the translation later.
Next, let’s choose a paragraph from an impromptu drama introduction for a translation test and compare it with Google and Youdao.
We can see that the differences are not significant at first, but the last paragraph shows a considerable difference. Overall, the content translated by DeepL is close to human semantics, reads accurately and fluently, and feels very natural, aligning better with human reading habits. It translates texts using trained artificial intelligence, attempting to add nuances and expressions of the language, not just sticking to literal translations, thus achieving exceptional translation results. This approach will allow DeepL to become increasingly accurate in the future, closer to human language habits.
2.2 Combining DeepL Translation System with Linguee Dictionary
The quick launch key for the DeepL translator on the PC client is Ctrl+C, and hotkey modification is supported. First, select the text to be translated, then click the designated area as shown in the picture or hold the Ctrl key and press C twice in succession to bring up the translation interface; the startup method is very fast and convenient.
It should be noted that there are differences in usage during English-to-Chinese and Chinese-to-English translations.
First, let’s take an example of a Chinese-to-English translation process:
When the translation result is in English, if certain translations are not precise enough, you can move the mouse directly to the word you want to select and click to see a series of alternative options as shown in the picture, and users can choose a more precise translation based on the context. However, when DeepL’s translation result is in Chinese, this functionality is not available, but there are corresponding solutions.
Next, let’s take an example of an English-to-Chinese translation process:
In the Chinese translation result, if we find that the word ‘revisit’ translated as ‘重温’ is not very accurate, it can be cumbersome to look it up in other platforms’ dictionaries. In fact, we can directly click on a word to see other translation results, for example, checking the translation results of ‘revisit’ in DeepL, which not only shows the basic verb meaning but also displays extended meanings such as ‘重访, 反思, 旧地重游’, allowing us to refine and polish the translation results based on these options.
The above operation process is illustrated with the local version. However, this functionality is available on both the web version and PC side, and the translation results are consistent.
2.3 Translating Entire Documents
The function of translating entire documents is available in both the free version of the software and the web version of DeepL Pro, and this function only supports a few file formats: Word (.docx), PowerPoint (.pptx), and txt documents. After translating the entire document, all fonts, images, and formatting remain unchanged, allowing users to edit the translated document according to personal preferences.
The document translation function is very intuitive and convenient to use. Hover the mouse pointer over the “Translate Document” button at the bottom left of the input box, upload the file, then select the target language to be translated into, and the translation will automatically begin, as shown in the picture below:
DeepL will automatically handle the remaining work, and the translation speed is very fast. The translated document can be downloaded locally and supports multiple downloads.
After downloading the web trial version of the translation results locally, the editing function is restricted; users can only view the translation results. Paid users can edit the downloaded files further.
Let’s roughly check the accuracy of the translation results:
It can be found that this segment of text has no differences from the previous web version and PC-side translation results, indicating that both the translation of the entire document and partial text should be consistent, and the accuracy has not been affected.
The advantage of this function is that it can translate large volumes of text in just a few seconds. For users who need to quickly browse foreign news and other materials, it is extremely convenient and fast. Moreover, DeepL’s official website states that more file formats will be supported in the coming months.
2.4 Explanation of Terminology Database and Other Extended Features
DeepL can also create custom glossaries to select the translation methods for words and phrases; this glossary will automatically tailor the user’s translation documents and texts. However, currently supported languages are only English and German, as well as English and French conversions. The official website states that more languages will be released soon.
DeepL Pro can also integrate with computer-assisted translation software (CAT Tool), allowing freelance translators, translation agencies, language service providers, or corporate language departments to use it in their CAT tools. Many leading CAT tools (such as SDL Trados Studio, memoQ, and Across Language Server) have plugins that integrate DeepL translation.
The following image shows the application of the DeepL Pro plugin in SDL Trados Studio 2017:
Finally, we list the comparison of services obtained from free and various paid levels, so everyone can judge according to their needs.
DeepL is an AI startup based in Germany that focuses on providing translation services using artificial intelligence. AI, machine learning, big data, neural networks, algorithms, and various keywords sound quite eye-catching, but for ordinary users, there is no need to understand the underlying principles too much. It is generally through a set of algorithms, combined with continuous learning and training of new translation data, that the accuracy of translations between different languages increases, and the results become increasingly natural. Many companies, both domestic and international, have similar AI translation services; foreign companies like Google Translate and Microsoft Translator are leaders in this field, while domestic companies like Tencent, Baidu, and NetEase also offer similar services, along with numerous startup projects. The effects of these mainstream translation software are not significantly different, and for users who need to quickly read foreign information, the choice of which to use does not make much difference. After all, if it’s just reading information, as long as the general meaning is translated, the human brain will automatically fill in and correct the understanding. However, if writing or composing foreign language emails is needed, it may be worth trying to use DeepL for initial translation and then modify and polish it, as it may be more natural in comparison.
DeepL’s translation seems to be moving towards the principles of fidelity, expressiveness, and elegance, and perhaps one day in the future, like AlphaGo in Go, a perfect machine translation will truly emerge.
Author: Xia Jingyi, Master’s in Translation, Dalian University of Foreign Languages
Reviewed by: Li Zhicheng, Zhu Yu, Shi Qinglin, Chen Mengna
This issue’s editor: Wang Xiaofan
