Document Retrieval by Hierarchical Multi Dimensional Scaling

通过分层多维缩放进行文档检索

• 批准号: 13680535
• 负责人: KANEKO Takeshi
• 金额: $ 1.92万
• 依托单位: Musashi Institute of Technology
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 2001
• 资助国家: 日本
• 起止时间: 2001 至 2002
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-13680535/
• 关键词: Knowledge Base; Multi Dimensional Scaling; Correspondence Analysis; Document Retrieval; Document Clustering; テキストデータ; 情報検索; 概念空間; クラスター分析; 自己組織化マップ; 意味空間

This research proposed the method ;1 constructing knowledge base which contain the information about relationship between words,2 retrieving distance information from the above knowledge base according to user view,3 document clustering based on the above distance information.Features of this method is ;1 Knowledge base is constructed by correspondence analysis that uses the frequency of words.2 Assessment of documents similarity is calculated on projection space by "view vector".The keywords user inputted estimates view vector.This method constructs good document clusters in some experiments. However, we can not constructs good document clusters when additional keywords are specified. When additional keywords are specified, re-build knowledge base gives preferable results.

本研究提出了方法；1 .构建包含词间关系信息的知识库，2 .根据用户视图从知识库中检索距离信息，3 .基于上述距离信息进行文档聚类。该方法的特点是；知识库是通过对应分析，利用词的频率来构建的文档相似度的评估是通过“视图向量”在投影空间上计算的。用户输入的关键词估计视图向量。在一些实验中，该方法构建了较好的文档聚类。然而，当指定了额外的关键字时，我们无法构建良好的文档集群。当指定了其他关键字时，重新构建知识库会得到更好的结果。