欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
利用汉-英双语句对进行了抽取短语翻译对的研究，提出了一种利用双语评价特征进行译文评价的短语翻译对主动获取方法。该方法通过选择有代表性的短语翻译对来达到减少人工标注数据的目的，以短语译文直译率、短语翻译概率和短语长度差异为基础，使用标注后的短语翻译对对支持向量机（SVM）进行训练，并使用优化后的SVM对测试数据进行分类。实验结果表明，使用此方法，在分类器性能基本没有下降的前提下，人工标注数据量减少了80％。
英文摘要：
In this study, phrase translation pairs were extracted from Chinese-English bilingual sentence pairs, and then, an active acquisition method for phrase translation pairs based on evaluation of phrase translation using bilingual evaluation features was proposed. The method achieves its purpose of decreasing manually-labeling cost by selection of representative phrase translation pairs, and uses annotated phrase translation pairs to train the support vector machine （SVM） based on the phrase translation literality, the phrase translation probability and the phrase length difference. And, it applies the optimized SVM to classification of test data. The experimental resuhs show that when using the method, the labeling cost can be reduced by at least 80 %, while the performance of the classifier remains constant.
同期刊论文项目
同项目期刊论文
期刊信息
位置：