语义关系下英语复杂长句机器翻译算法优化

Optimization of machine translation algorithm for English complex long sentences based on semantic relations

传统机器翻译算法对英语复杂长句的翻译准确率低、回收率高,针对这一难题,提出了基于语义关系的机器翻译算法。该算法通过长句内分句(短语)之间的语义层次关系,构建基于相似度的语义网络模型,结合余弦相似度和带权向量加法计算获得翻译结果,利用权重训练获得关键短语。以NIST 06和NIST 08测试集为例,基于语义关系的英语复杂长句机器翻译算法测试BLEU值比传统算法分别提高了0.35和0.23,即翻译结果的准确率提高,回收率降低。

Aiming at the low accuracy and high recovery rate for the translation of complex long sentences from English in traditional machine translation algorithms,it proposes a machine translation algorithm based on semantic relations.The algorithm constructs a semantic network model of the translation similarity model through the semantic hierarchical relationship within the long sentence,combines the cosine similarity and the weighted vector addition calculation to obtain the translation result,and uses the weight training to obtain the key phrase.The results show that the semantic relationship-based machine translation algorithm for English complex long sentences can optimize the translation results,and the test BLEU value increases by 0.35 and 0.23,that is,the accuracy of the translation results is improved,and the recovery rate is reduced.