神经网络机器翻译问题及对策研究——以有道、DeepL、搜狗、百度翻译为中心

Research on Neural Network Machine Translation Issues and Strategies—Focusing on Youdao, DeepL, Sogou, and Baidu Translation

摘要: 2013年以来，神经网络机器翻译实现了机器翻译质量的大幅提升，短短几年便成为机器翻译系统的主流模式。国内外行业巨头纷纷上线神经网络机器翻译系统，推动机器翻译实用化和商品化发展。但是，神经网络机器翻译仍然面临诸多挑战。如何有效地利用在线翻译系统，方便、快速且准确地提高翻译效率和质量，是当前机器翻译研究的重点。因此，本文以有道、DeepL、百度、搜狗四个神经网络翻译软件为研究对象，概括总结了神经机器翻译译文的特点及存在问题，并据此从机器翻译软件开发、译后编辑、译者发展等不同角度，探讨在人工智能翻译飞速发展背景下的应对策略。

Abstract: Since 2013, neural network machine translation has significantly improved the quality of translation and has quickly become the mainstream mode of machine translation systems. Industry giants have launched neural network machine translation systems, promoting the practical and commercial development of machine translation. However, neural network machine translation still faces many challenges. How to effectively use online translation system to improve the efficiency and quality of translation conveniently, quickly and accurately is the focus of current machine translation research. Therefore, this paper takes Youdao, DeepL, Baidu, and Sogou as the research objects of neural network translation software, summarizes the characteristics and existing problems of neural machine translation, and discusses the coping strategies under the background of rapid development of artificial intelligence translation from different perspectives such as machine translation software development, post-translation editing and translator development.