Web version screen of DeepL translation. An app version is also available.
I had the opportunity to interview Mr. Jaroslaw Kutylowski, the CEO of DeepL, who was visiting Japan.
DeepL’s feature is “translation using AI”, but how is it currently being used? And how do you see the future of translation? We asked Mr. Kterowski, who is also an engineer, about his vision.
Security and “confidentiality” created DeepL’s corporate penetration
Jaroslaw Kterowski, CEO of DeepL.
First, let’s review the history of DeepL.
DeepL is a German company that provides a translation service called DeepL Translation. Headquarters are in Cologne. Since its launch in 2017, it has grown into a service used around the world. In particular, after supporting Japanese in 2020, it has become a widely used service in Japan.
*The company name was changed from the original “Linguee” to the current “DeepL” in 2017.
There are many competitors in automatic translation. Big companies like Google have long offered translation services for free.
However, when DeepL appeared and Japanese support was announced, it began to become a hot topic in Japan. The reason is, of course, quality.
“I think that the quality and accuracy of the translation is what makes it possible. As you continue to use it, you will become more confident that ‘this will work for you,’ so the tool will become more trustworthy.
DeepL seems to have developed such a relationship of trust with its users from an early stage.” (CEO Kterowski)
Another big point is the business model.
“especiallySecurity is important for large companiesIt is also important that it was They routinely exchange documents that require confidentiality. Having a ‘Pro’ plan with security and confidentiality guarantees was an important point in gaining their trust.” (CEO Kterowski)
In cloud-based and AI-powered translation, data is often collected and used for learning.
However, if you have a confidentiality obligation, you can not use a service that may disclose the contents of the document to other companies. But while DeepL has a free “freemium” model, the paid “Pro” plan clearly states that any document is “immediately destroyed after translation,” making it easier for businesses to use.
Paying users are “more than 500,000 worldwide”says Couteroski. There are more than tens of millions of free users, but it can be said that the number is not low for a freemium model.
But Kterowski also said, “We try not to refer to free users as consumers.”
“Some people use DeepL for private purposes such as communicating with their families, but most people use it for some kind of business scene. ,” (CEO Kterowski)
Advertisements
College student “write a thesis with DeepL”.The writing style has changed
As more people started using services like DeepL, an interesting phenomenon began to emerge.
When a student writes a thesis, he first translates what he wrote in Japanese into English with DeepL, and then translates that English into Japanese with DeepL.I am improving difficult and hard-to-understand phrases and learning how to express myself in English.That’s what it means.
In other words, by writing sentences that are easy for AI to translate, we aim to create sentences that are easy to understand even in other languages.
“It’s an interesting trend.DeepL is not intended for language learning, but more and more people around the world are using it that way.In Germany there are people translating from German to English and from English to German.
I myself sometimes use this method when translating something written in German into English. When I didn’t like the English translation, I realized, ‘Oh, my writing wasn’t clear,'” says Kterowski.
Kterowski is fluent in English as well as his native German, but says he still uses DeepL on a daily basis.
“Perhaps there is a difference in how DeepL is used in Japan and Germany. In Germany, it is important to focus on efficiency and to do things very quickly.
For me, it’s sometimes easier and faster to write something in German and translate it right away. You can translate it yourself or write it directly in English, but it’s important to finish it quickly. Especially in business, people always want to save time.”
This feeling is very familiar. In fact, I use it as well.
“Of course, there are a lot of people who use DeepL for languages they can’t read or speak,” said Kterowski. DeepL will often be used to reduce
Does the amount of trade affect the quality of AI translation?
Accuracy comparison with other translation tools published by DeepL.
In terms of translation, there are two challenges ahead.
The first is to improve the accuracy of the translation. In AI translation, it is said that “the more speakers of a language, the more information it has and the more advantage it has”. In that case, English and Chinese are overwhelmingly advantageous, but what about that point?
“To be precise, it is a problem of the amount of translation data that can be searched on the Internet etc.I think.
It doesn’t necessarily have to do with the amount of people who speak the language. For example, if you analyze bilingual data on the Internet, you will find many German-English texts. This is because Germany exports a lot.
I have heard that the Tokyo Stock Exchange recommends publishing quarterly financial results in both Japanese and English. This can be a very useful source of additional translation data.
in short,A country’s openness to its language and the amount of trade influences the amount of translationI think that’s why.” (CEO Kterowski)
On the other hand, of course, it is also important to reduce the dependence on the amount of data. However, in order to improve the quality of translation, another perspective is also necessary.
“I think we’re headed in the direction of getting better and better at translation without having a huge amount of data. But if you want translation and translation technology to work on different topics, you have to learn about the topics themselves. is needed.
If you want to translate medical texts or microbiological texts, an understanding of terminology and writing style is essential.
We want AI to have the capabilities of many human translators, specialized in specific contexts. It’s a huge challenge for AI and requires a lot of data.”
There are also mistranslations in the current DeepL. Therefore, it is sometimes said that “DeepL should be used by people who understand English.” This is especially true because DeepL can skip paragraphs of complex phrases, changing the meaning of the sentence.
What is the solution to this problem?
“We are aware of the issue and will definitely pass it on to our AI development team to make improvements.”
“Of course, we have to improve,” says Kterowski with a wry smile.
“This is a common problem with AI and neural networks working. Sometimes we skip the difficult parts, just like us humans[bitter smile]” (CEO Kterowski)
(Written by Sochika Nishida)
Source: BusinessInsider
Emma Warren is a well-known author and market analyst who writes for 24 news breaker. She is an expert in her field and her articles provide readers with insightful and informative analysis on the latest market trends and developments. With a keen understanding of the economy and a talent for explaining complex issues in an easy-to-understand manner, Emma’s writing is a must-read for anyone interested in staying up-to-date on the latest market news.