基于动态词对齐的交互式机器翻译-中科院智能农业信息网.pdfVIP

第十一届全国机器翻译研讨会(CWMT 2015) 中国 · 合肥 2015.9.24-25 基于动态词对齐的交互式机器翻译 马斌，蔡东风，季铎，叶娜，吴闯 （沈阳航空航天大学 人机智能研究中心，辽宁 沈阳 110136） 摘要： 在传统的机器翻译（Machine Translation，MT）与计算机辅助翻译（Computer Aided Translation， CAT）中，译员与翻译引擎之间的交互受到很大限制，于是产生了交互式机器翻译（Interactive Machine Translation，IMT）技术。但传统的模型只考虑当前源语与部分翻译的目标语的信息，没有将用户交互后 的对齐信息加入到之后的预测模型中。本文基于词预测交互式机器翻译的研究思路，将用户交互翻译过程 中的鼠标点选行为转化为中间译文的词对齐信息，进而在翻译交互过程中实现了对译文的动态词对齐标注， 并在词对齐信息和输入译文的约束下提高了传统词预测的准确性。 关键词：交互式机器翻译；词对齐；预测模型 中图分类号：TP391 文献标识码：A Improving Interactive Machine Translation by Dynamic Word Alignment MA Bin, CAI Dongfeng, JI Duo, YE Na, WU Chuang Human-Computer Intelligence Research Center , Shenyang Aerospace University , Shenyang , Liaoning （ 110136 , China ） Abstract: In traditional machine translation (MT) and computer aided translation (CAT), the interaction between human translators and machines is very limited. So the interactive machine translation (IMT) technology has emerged. The traditional models didn ’t add the alignment information obtained in the interaction of translators and machine into the subsequent prediction models and only considered the information of the current source language and the partial translation of the target language. In this paper, we propose a method based on the research of the word prediction interactive machine translation, translate the selection behavior of the mouse in the interactive process into the word alignment information of the partial translation, and then label the dynamic word alignment of the translation. The approach has improved the accuracy of prediction under the constraint of the dynamic word alignment and the input translation. Key words: Interactive Machine