一种基于词语搭配的语义消歧方法

提出一种利用规则与统计相结合的方法用于英汉机译系统中以消解语义级歧义 ,建立了一种根据单词之间的词语搭配关系以消除歧义的模型。该模型利用英汉双语语料选择合理的词组语义 ,对有歧义的单词作出标注。在此基础上给出了语义消歧的学习算法 ,并建立了一套有效的提高召回率的消歧算法。算法在英汉机译系统中的实际应用使正确率提高了约 10 % ,效果显著。

Word sense tagging method based on word relationships

The paper proposes a word sense tagging method using a hybrid strategy combining rule based and statistic based methods. A mathematical model is constructed to remove the word ambiguity in term of its relationship to other words. By evaluating each possible sense combination according to bilingual corpus, the model can select the most reasonable word sense. Then, a rules learning algorithm is proposed and an application algorithm of the learned rules is provided which can increase recall. Its application in English Chinese machine translation system proved a 10% increase in precision.