基于矩阵加权关联模式的印尼中跨语言信息检索模型

发布时间：2018-05-18 01:29

本文选题：点击行为 + 关联模式挖掘　；参考：《数据分析与知识发现》2017年01期

【摘要】：【目的】针对跨语言信息检索存在的查询漂移问题,提出一种融合用户点击下载行为与矩阵加权关联模式挖掘的印尼中跨语言信息检索模型。【方法】将矩阵加权关联模式挖掘、查询扩展以及用户点击下载行为集成应用到印尼中跨语言信息检索模型,给出模型实现的关键技术,即面向跨语言信息检索的矩阵加权关联模式挖掘算法、跨语言查询扩展模型以及印尼中跨语言信息检索算法。【结果】在NTCIR-5 CLIR数据集上的实验结果表明,该检索模型的R_prec、p@10和p@20值均达到单语言检索基准的60%以上,比跨语言检索基准提高37%以上,比现有基于伪相关反馈的跨语言检索算法提高28%以上。【局限】该模型实验在基于向量空间模型的跨语言检索系统中进行,需要探讨和研究在实际搜索引擎中的具体应用。【结论】该模型能有效地减少跨语言检索中的查询漂移问题,提高和改善印尼中跨语言检索性能,对长查询的检索效果更好,有较好的实际应用价值。
[Abstract]:[objective] to solve the problem of query drift in cross-language information retrieval, This paper proposes a cross-language information retrieval model in Indonesia, which combines user click-download behavior with matrix weighted association pattern mining. [methods] Matrix weighted association pattern mining is proposed. Query expansion and user click-download behavior integration are applied to the cross-language information retrieval model in Indonesia. The key technologies of the model are given, which is matrix weighted association pattern mining algorithm for cross-language information retrieval. Results the experimental results on the NTCIR-5 CLIR data set show that the RSP @ 10 and PDP 20 values of the retrieval model reach more than 60% of the monolingual retrieval benchmark. It is more than 37% higher than the cross-language retrieval benchmark and more than 28% higher than the existing cross-language retrieval algorithms based on pseudo-correlation feedback. [limitation] this model experiment is carried out in a cross-language retrieval system based on vector space model. This model can effectively reduce the query drift in cross-language retrieval, improve and improve the performance of Indonesian mid-to-medium cross-language retrieval, and improve the retrieval effect of long query, which needs to be explored and studied in practical search engine. [conclusion] this model can effectively reduce the query drift in cross-language retrieval, and improve the performance of mid-to-medium cross-language retrieval in Indonesia. It has good practical application value.
【作者单位】：广西跨境电商智能信息处理重点实验室培育基地(广西财经学院);广西财经学院计算机系;
【基金】：国家自然科学基金项目“面向东盟国家语言的基于完全加权正负模式挖掘的跨语言查询扩展研究”(项目编号:61262028) 广西财经学院信息与统计学院开放性课题“基于矩阵加权关联模式挖掘的越汉英跨语言信息检索研究”(项目编号:2015XK01) 广西财经学院2016年度应用统计硕士专业学位点学术研究项目“基于完全加权关联模式挖掘的中英跨语言伪相关反馈扩展研究”(项目编号:2016TJYB05)的研究成果之一
【分类号】：TP391.3

【参考文献】