【摘要】 介绍了使用混合策略从汉英双语语料库中抽取搭配的方法。采用互信息抽取最初的候选的搭配组合,并用t-测试值考察其可靠度,过滤掉t-score<1.65的候选搭配串,再通过词性标注和浅层句法分析进行筛选。实验证明了该方法的有效性。同时探讨了将抽取的搭配组合应用于建造双语词典和机器翻译系统的途径。更多还原

【Abstract】 This paper introduces a new approach to extract collocation from Chinese-English parallel corpora using hybrid strategy.We extract initial collocation candidates using mutual information and calculate t-score to eliminate those below the threshold of1.65.Then the candidate collocations left are further filtered through POS patterns and chunking.Relative Experiments prove the validity of this method.Finally,we show the ways to apply the results to bilingual dictionary construction and machine translation system.更多还原