On February 21st, the China Translation Association and the Translation Institute of the China Foreign Languages Administration held a seminar on empowering translation work with artificial intelligence. The conference will be held in a combination of online and offline formats, focusing on the technological breakthroughs of the domestic large-scale model DeepSeek and its profound impact on the translation industry. Industry experts will be invited to conduct in-depth discussions on technological development trends, industry transformation paths, and talent cultivation directions. Gao Anming, Chief Editor of the China Foreign Languages Administration and Executive Vice President and Secretary General of the China Translation Association, attended the meeting and delivered a speech. Experts from China Internet News Center, Beijing Language and Culture University, Guangdong University of Foreign Studies, Dalian Foreign Studies University, Huawei, Yizhe Technology, Transversal, Language Bridge, Oracle Bone Yi, Shenzhen Yunyi, Chengdu Youyi and other units attended the meeting.
Gao Anming pointed out that the development and application of artificial intelligence have brought new opportunities and challenges to the translation industry. He stated that high-quality data, expert feedback, and application scenario requirements can help optimize and upgrade artificial intelligence technology services, and translation and international communication experts should be more involved in the iterative loop of artificial intelligence models. He emphasized that the translation industry will be an industry where human translators and artificial intelligence complement and interact, and develop together. The China Translation Association will further play a leading role in the industry, strengthen cooperation and co construction with all parties, improve platforms, standardize applications, and collaborate, deepen exploration and practice of the integration of translation, communication, and technology development, and promote the development of new quality translation productivity through practical measures.
The Rise of Domestic Large Models: DeepSeek Technology Advantages and Application Challenges Coexist
Experts attending the conference believe that DeepSeek has become a leading global model due to its open-source properties, reasoning capabilities, and cost advantages.
Ding Li, Chairman of Shenzhen Yunyi Technology Co., Ltd., pointed out that DeepSeek performs better than some mainstream models in English Chinese translation tests, performs well in multiple language translations, and has surprising translation effects in professional fields such as medicine. Especially after combining terminology and memory, the initial translation effect has significantly improved, but there are still accuracy issues that require deep manual involvement in the post translation editing process.
Han Lintao, Vice Dean of the School of English and Advanced Translation at Beijing Language and Culture University, emphasized the importance of distinguishing the characteristics of different versions when exploring model performance. He pointed out that preliminary experimental results in specific fields between Chinese and English indicate that DeepSeek V3 is comparable to GPT-4o in terms of translation ability; In terms of terminology extraction ability, DeepSeek V3's ability to extract Chinese political terminology is slightly inferior to GPT-4o; In terms of auxiliary programming ability, DeepSeek V3 is weaker than Claude 3.5 Sonnet. In terms of reasoning ability, by combining translation memory and terminology, the locally deployed DeepSeek R1 can further improve translation quality. However, currently DeepSeek still faces issues such as traffic limitations and slow API call speeds.
Xie Ning, the head of machine translation products and data at Huawei Technologies Co., Ltd., analyzed from the perspective of architecture innovation and pointed out that DeepSeek will accelerate the development of ultra large parameter MoE models and inference models, and bring about more profound changes in translation modes and processes.
Wang Shaoshuang, Vice Dean of the School of Advanced Translation at Dalian Foreign Studies University, believes that DeepSeek's R1 model has outstanding reasoning ability and outputs language that is more "human like". He pointed out that DeepSeek includes both general models and inference models, and in practical applications, suitable models need to be selected based on the type of task.
Wei Tianbing, General Manager of Jiaguyi (Beijing) Language Technology Co., Ltd. Shandong Branch, emphasized the advantages of DeepSeek as a localized product, such as ease of use, low cost, and local deployment to avoid data leakage. At the same time, he also pointed out that the model has problems such as slow inference speed and possible weakening of cultural characteristics.
New paradigm of human-machine collaborative translation: pre translation and post translation editing into key battlefields
In terms of translation research, experts believe that artificial intelligence provides researchers with richer data resources and analysis tools. Yan Lili, Vice President of Transsion Network Technology Co., Ltd., pointed out that with the help of artificial intelligence, researchers can conduct regional and country specific research more efficiently, promoting the deeper development of translation research. Zhang Jing, General Manager of Shanghai Yiyi Information Technology Co., Ltd., believes that the deep integration of DeepSeek and knowledge base can significantly improve translation quality and will also play a key role in assisting corpus research.
In terms of industry applications, experts unanimously believe that artificial intelligence has brought unprecedented opportunities and challenges to the translation industry. Wei Tianbing mentioned that DeepSeek's strong ability to translate between Chinese and English may bring significant changes to the interpreting industry. Ding Li pointed out that the application of artificial intelligence models such as DeepSeek not only improves translation efficiency, but also reduces costs, enabling more translation companies to provide more cost-effective translation services.
In terms of translation mode, experts believe that pre translation and post translation editing have become key battlefields in translation work. Zhao Xuan, director of the Technology Innovation Department of the China Internet News Center, pointed out that making full use of DeepSeek's understanding of Chinese to edit the original text before translation can effectively improve the effective transformation of Chinese content into international communication. Wang Weiwei, Vice Dean of the School of Advanced Translation at Guangdong University of Foreign Studies, believes that there are "hidden errors" in AI generated translations, such as concept substitution and cultural misreading. Deep problems such as terminology consistency and cultural logic conversion still need to be checked by professional translators, and an industry level quality review mechanism needs to be established, as well as the development of AI translation risk grading usage standards based on application scenarios.
Yan Lili frankly stated that while the initial translation cost has significantly decreased, post translation editing must be completed by professional and versatile translation experts. Wei Tianbing stated that the illusion problem of large models may lead to errors and omissions in the generated content, which puts higher demands on translators' ability to distinguish the authenticity of AI output content. Zhu Xianchao, Chairman of Sichuan Language Bridge Information Technology Co., Ltd., pointed out that a single large model product cannot perfectly adapt to all translation tasks, and the role of experts is still the most important in translation quality management.
Reshaping Industry Ecology: Open Source Technology Accelerates Intelligent Transformation
The emergence of artificial intelligence models such as DeepSeek has not only changed the way translation research and practice are conducted, but has also had a profound impact on the translation industry ecosystem.
Ding Li pointed out that the popularization of technology has had an impact on traditional translation business models, prompting translation companies to reposition their business models and business strategies.
Zhang Macheng, CEO of Chengdu Youyi Information Technology Co., Ltd., pointed out that the advantages of high-quality translations, wide brand dissemination, and excellent user experience of large models have reduced users' demand for translation companies, which has had a negative impact on translation companies to a certain extent. The business volume has decreased year-on-year, and companies should actively take corresponding measures.
Zhang Jing believes that as global competition for large-scale models intensifies, the improvement of inference efficiency and cost reduction will drive their comprehensive application in translation and other fields. DeepSeek has gradually become an infrastructure capability, integrated into the workflow of the public. The development of the translation management platform can be upgraded and iterated with the help of the DeepSeek big model, and intelligent agent technology will promote the automation of translation project processes and be fully implemented.
Experts unanimously believe that with the advancement of artificial intelligence technology and the expansion of application scenarios, translation companies will gradually shift from providing traditional translation services to providing diversified and intelligent services such as content services and international communication, in order to meet the increasingly diverse needs of customers.
Translation Education Innovation: Deep Integration of Humanities and Technology
Faced with the impact of artificial intelligence models such as DeepSeek, the training of translation talents in the new era is also facing new requirements and challenges. Experts believe that in the new era, translation talents not only need to have a solid language foundation and professional knowledge, but also need to have advanced technology and tool application abilities. At the same time, universities should actively adjust their training objectives and curriculum settings, strengthen cross disciplinary integration with majors such as computer science and information technology, and cultivate more versatile translation talents that meet the needs of the artificial intelligence era.
Zhang Macheng found that high-level translators are better at detecting and correcting artificial intelligence errors, while low-level practitioners may be eliminated by technology. Yan Lili suggests that universities use artificial intelligence tools to shorten the language learning cycle, strengthen students' composite competitiveness in "professional fields+foreign language abilities", and accelerate the cultivation of artificial intelligence translation talents in non common languages and ethnic languages. Wang Weiwei believes that translation education urgently needs to promote personalized, layered, and precise teaching models, implement a formative evaluation system, and cultivate students' ability to critically use artificial intelligence translation tools. At the same time, the translation industry should accelerate the construction of an AI translation ethics review mechanism. Wang Shaoshuang emphasized that attention should be paid to cultivating the intelligent translation literacy and translator prompt literacy of translation majors, and translation teachers should continuously improve their translation practice and teaching abilities.
Looking forward to future development: Embracing technology, adhering to professionalism, and inheriting mission
When looking forward to the prospects of the integration of translation technology, experts emphasized the importance of continuous innovation and cooperation. It is believed that although artificial intelligence models such as DeepSeek have achieved significant results in the field of translation, there are still many areas that need improvement and refinement. Language experts and technical experts should strengthen communication and collaboration to jointly promote the upgrading and optimization of translation technology.
Xie Ning pointed out that DeepSeek's technical architecture and training methods provide a new direction for translation technology research and development. In the future, translation technology research and development should pay more attention to the deep integration with artificial intelligence technology, and explore more efficient and intelligent translation solutions. Zhao Xuan emphasized the potential of DeepSeek in multilingual content generation and cultural transformation from an international communication perspective. She called on the translation industry to actively explore the integration and development of artificial intelligence technology to enhance international communication efficiency and cultural influence.
Looking ahead to the future, with the continuous advancement of technology and the continuous expansion of scenarios, artificial intelligence technology will undoubtedly serve the needs and development of the translation industry more intelligently and personalized. All parties in the translation industry should actively embrace changes, strengthen cooperation, continuously innovate and upgrade service capabilities and models, and use high-quality language services to promote the construction of the external discourse system.
This seminar gathered the insights of senior experts in translation practice and technology research and development, delving into the multifaceted impact of DeepSeek on the translation industry. It provided many ideas and useful references for exploring the collaborative development model of artificial intelligence and the translation industry, and promoting the construction of national translation and international communication capabilities. As experts have pointed out, the big model has no end, only by maintaining inclusiveness can it stand undefeated in this "intellectual revolution".
The seminar was chaired by Jiang Yonggang, Director of the Eurasian Center of the China Foreign Languages Administration, Vice President of the China Translation Association, and Chairman of the Translation Technology Committee. Officials from departments and affiliated units of the China Foreign Languages Bureau attended the meeting.