By Kim Boram
SEOUL, May 11 (Yonhap) -- Jaroslaw Kutylowski, founder and CEO of the neural machine translation service provider DeepL, has said demand for Korean language on the global stage is growing larger and faster amid the rising presence of the world's fourth-largest economy in the time of artificial intelligence (AI)-led digitalization.
"There is going to be a lot of demand for Korean language," the businessman said in an interview with Yonhap News Agency held in Seoul on Tuesday.
"I think it is a country that is technologically very advanced and therefore will adapt these kinds of solutions very well," he said. "It has a lot of interaction with other countries not only in the Asian region, but also in the U.S. and Europe."
DeepL is a Germany-based AI company founded in 2017 that provides instant translation-as-a-service both to businesses and to individuals. Its translation service supports 31 languages, including English, German and Chinese.
It launched the Korean service in early January, and its paid, tailored service, DeepL Pro, will be released in August.
Kutylowski, who had been chief technology officer of DeepL from the beginning, said he has chosen Korean as the company's 31st language, and the fourth Asian language after Chinese, Japanese and Indonesian.
He said it took time for the company to develop technologies for the Korean language, which has different grammar and usage, especially honorifics and its alphabet, hangeul, from Western languages, like English and German.
Out of the 31 language offerings, the other 27 are European languages.
"It's like pieces of advanced technology. It's the sum of little things. That is complicating to get there," he said. "The Korean language system, especially the writing system, is a very specific one. So this is also something that we needed to get over with."
DeepL is using an artificial neural network for its translation that predicts the likelihood of a sequence of words and models entire sentences based on machine learning of billions of documents and materials.
The official, who has a German and Polish background, said the quality of translation by DeepL depends on how much its AI has learned from available data in certain languages.
"The larger those data sets are given in general, the better the quality of its generative AI or translation we have," he said. "It's just a fact that explains why German-English translation models are very good. it's because there's a lot of training data for that."
He admitted that an AI program needs a lot of data for training before producing a product and he is now working on the training data issue, which needs a wide range of data collections of a language before launching a service in a new country, including South Korea.
"It will be definitely a research goal in the future to train models on smaller data sets and for them to be able to learn from less data," he said. "We had quite a lot of complicating works to gather data on the Korean internet.
He said he has plans to open a regional office in South Korea if its Korean customers want, after sale service in time.
Still, DeepL has many Korean-speaking officials at its German headquarters to provide quality Korean service for customers here, he added.
"It depends on how the demand for DeepL Pro is going to be in Korea and what kind of approach Korean companies will require from DeepL as our potential customers," he said. "We think, will they need somebody in Korea to help them, or will we be able to work with this market from Europe or from other countries, like the U.S."