TapTechNews July 18th news, the language AI company DeepL, known for its namesake machine translator, announced yesterday local time in Germany the launch of a new generation of large language models for translation and editing applications.
DeepL claims that the translated texts based on the new model are favored by language experts by 1.3 times that of Google Translate, 1.7 times that of ChatGPT-4, and 2.3 times that of Microsoft Translator.
The new model of DeepL is based on the training corpora accumulated by this enterprise in the past seven years, adopts advanced large language model technology, and has received the 'tutoring' of thousands of language experts, specifically designed for content creation and translation.
DeepL points out that when performing English-Japanese and English-Chinese mutual translation, the new model has significantly improved by 1.7 times in quality compared to the old model; in addition, the quality of English-German mutual translation has also increased by 1.4 times.
This means that the new model can greatly reduce the manual re-editing after machine translation, improving productivity for enterprises that need to translate a large number of documents: to achieve the same quality as the new model of DeepL, Google Translate requires twice the amount of editing, while ChatGPT requires three times the amount of re-editing work.
TapTechNews learned that DeepLPro paid subscription users can now use the new generation model for mutual translation between English and Japanese, German, and simplified Chinese, and these users can switch between the old and new models to obtain a more flexible translation experience.