融合知识图谱的查询扩展模型及其稳定性研究

发布时间：2018-04-05 11:45

本文选题：知识图谱　切入点：Freebase　出处：《计算机科学与探索》2017年01期

【摘要】：旨在构建一种基于知识图谱Freebase的查询扩展模型,通过从Freebase中抽取与查询相关的若干实体及实体属性作为扩展词来重构查询,从而更好地表达用户的信息需求。在计算扩展词权重时,参考了投资组合理论中收益-风险分析方法,最大化扩展词和查询的相关性收益,同时也最小化扩展词可能带来的查询漂移的风险。最后将查询相关的实体和实体属性作为两种特征和查询语言模型结合实现查询扩展。在两个Web数据集上进行实验,用来检验所提出的扩展模型对检索系统的有效性和稳定性的影响。实验结果表明,提出的查询扩展模型与一元语言模型LM相比,检索结果的平均准确率(mean average precision,MAP)在两个数据集上有6%至15%的显著提升;和基于伪相关反馈的查询扩展模型RM3相比,有效性及稳定性都有不同程度的提升。
[Abstract]:This paper aims to construct a query extension model based on knowledge Map (Freebase). By extracting some entities and entity attributes related to the query from Freebase as extension words, the query can be reconstructed to better express the information needs of users.In calculating the weight of extended words, we refer to the profit-risk analysis method in portfolio theory to maximize the correlation benefit between the extended words and the query, and minimize the risk of query drift caused by the extended words.Finally, query extension is implemented by combining query related entities and entity attributes as two features and query language models.Experiments were carried out on two Web datasets to test the effects of the proposed extended model on the effectiveness and stability of the retrieval system.Effectiveness and stability are improved to varying degrees.
【作者单位】：天津大学天津市认知计算与应用重点实验室;
【基金】：国家自然科学基金国家重点基础研究发展计划(973计划) 高等学校博士学科点专项科研基金~~
【分类号】：TP391.3

【相似文献】