摘要
词义消歧一直是自然语言理解中的一个关键问题,该问题解决的好坏直接影响到自然语言处理中诸多问题的解决。现在大部分的词义消歧方法都是在分词的基础上做的。借鉴前人的向量空间模型运用统计的方法,提出了不用直接分词而在术语抽取的基础上做消歧工作。在义项矩阵的计算中,采用改进了的tf.idf.ig方法。在8个汉语高频多义次的测试中取得了平均准确率为84.52%的较好的效果,验证了该方法的有效性。
Word sense disambiguation is a key problem in natural language processing because the result of WSD affects seriously many problems in natural language processing. Most of word sense disambiguation are based on the segment of words. An unsupervised approach based on term extraction instead of segment of words is presented. The method of tf.idf.if is adapted in calculating the matrix of word sense. The result of average accuracy 84.52 % is gotten in the experiments on 8 ambiguous words, so the method has high value in research.
出处
《计算机工程与设计》
CSCD
北大核心
2007年第5期1215-1218,共4页
Computer Engineering and Design
基金
广东省自然科学基金项目(04105385)
