跨语言查询扩展优化  

Optimization of cross-language query expansion

作　　者：魏露[1] 李书琴[1] 李伟男[1] 李新乐[1] 

机构地区：[1]西北农林科技大学信息工程学院,陕西杨凌712100

出　　处：《计算机工程与设计》2014年第8期2785-2788,2803,共5页Computer Engineering and Design

基　　金："十二五"国家科技支撑计划基金项目(2012BAH30F01;2011BAD21B05;2013BAD15B02);技术转移集成服务平台建设与示范基金项目(2012BAH30F01);中央高校基本科研业务费基金项目(QN2011036)

摘　　要：为提高跨语言查询扩展检索精度,在原有跨语言查询扩展基础上,引入降低噪声和孤立点的k-medoid聚类算法,提出避免语义信息丢失或过拟合的择优模型。构建若干个不同维度值的d维模型,结合奇异值分解和非负矩阵分解法计算文本之间的相似度,选取相似度最大的模型建立双语空间,经过跨语言扩展与权值调整,实现查询扩展优化。实验对比结果表明,该方案有效提高了检索精度,为跨语言查询提供了可参考的模型与算法。To improve the cross-language query expansion retrieval precision, based on the original cross-language query expansion, the k-medoid clustering algorithm was introduced to reduce the noise and isolated points, the merit-based model was proposed to avoid the semantic information missing or over-fitting. A number of ddimensional models were constructed with different dimension values combined with the singular value decomposition and the nonnegative matrix factorization method to calculate the similarity between the texts, the bilingual model space with the greatest similarity was selected, through cross-language extensions and weights adjustments, finally the query expansion optimization was achieved. Experimental comparison results show that the improved scheme can improve the retrieval accuracy effectively, and it provides a reference model and algorithm queries for the cross-language field.

关 键 词：潜在语义 跨语言 查询扩展 k-中心点聚类 非负矩阵分解 择优模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]