人机协同翻译方法及关键技术研究

在全球经济一体化急速加快和创新型国家的建设进程中，我国对先进标准、专利、规范资料等海量工程文献的翻译需求日益迫切。尽管以深度学习技术为核心的神经翻译方法使机器翻译实用化方面取得进展，但其仍然难以满足专业技术资料大规模、高质量翻译和译文可维护的普遍需求，面临领域知识支撑、翻译任务长期变更维护和整体质量控制等诸多挑战。在混合智能理论迅速发展的背景下，对人机协同翻译方法及其关键技术的研究成为突破海量工程文献翻译的高效和高质量难题的必要途径，具有重要的理论和应用价值。人机协同翻译研究的关键是基于知识工程和知识管理的思想，针对当前主流机器翻译方法面临的知识支撑、模型优化和系统可解释性等瓶颈问题，探索面向人机协同翻译的知识图谱构建方法、人机协同翻译模型及其优化方法以及翻译知识场景化应用方法，构建以人为本、以知识为核心、以服务场景为驱动的协同翻译模式，在翻译全过程实现人与知识、人与模型、人与系统的协同。

In the age of rapid global economic integration and innovative country construction, our country is in urgent need of the translation of advanced standards, patents, specifications and massive technical documents. Although the neural machine translation method based on deep learning technology has promoted the application of machine translation technology, it is still difficult to meet the general needs of large-scale, high-quality and maintainable professional technical documents, and it faces many challenges such as domain knowledge support, long-term change maintenance and overall quality control. Under the background of the rapid development of hybrid intelligence theory, the research on Human-Machine Cooperative Translation method and its key technologies has become a necessary way to solve the problem of high efficiency and high quality in massive engineering documents translation, which has great theoretical and practical value. The key to the study of Human-Machine Cooperative Translation is exploring the construction of knowledge graph for translation, the Human-Machine Cooperative Translation model and its optimization method, as well as the application method of translation knowledge and scenarios based on the idea of knowledge engineering and knowledge management, so as to solve the knowledge bottleneck, the model optimization bottleneck and the system interpretability bottleneck of state-of-the-art machine translation methods, and construct a human-oriented, knowledge-centered, service scenario-driven collaborative translation mode which realizes human-knowledge cooperation, human-model cooperation, and human-system cooperation in the whole translation process.