机器翻译方法研究与发展综述

简要述评了机器翻译方法，依据知识处理方式，将其划分为三类，分别为规则法、语料库法、混合法，对比了其优劣所在。捕捉了当前机器翻译主流构建方式，基于注意力的神经网络法，其最新动态主要包括降低计算复杂度，提高词对齐质量，融入先验与约束，改善历史记忆力。基于上述考察指明未来机器翻译方法的发展趋势，具体包括网络融合逐步深化，处理方式走向并行，训练方法日趋多样。

Review of Studies and Developments on Machine Translation Methodology

First of all, machine translation methods are streamlined into three types based on their ways to knowledge processing, i.e. rule-based methods, corpus-based methods, and hybrid methods, and their merits and demerits are compared. Then, the latest study trends mainly comprise computational complexity reducing, words alignment enhancing, prior knowledge and constraints incorporating, and historical memory improving, attention-based neural networks, as the current mainstream methods of machine translation, are captured. Finally, the future development orientations are prospected, particularly, networks integration deepening, data processing parallelizing and training methods diversifying of machine translation methods.