基于特征词权重变更的检索优化策略  

Optimization Strategy of Information Retrieval Based on Updating Weight of Feature Words

作　　者：黄思思[1] 

机构地区：[1]华东师范大学工商管理学院,上海200241

出　　处：《情报科学》2016年第7期70-75,共6页Information Science

摘　　要：以无限逼近用户的真实检索需求为目的,基于传统检索系统的运行机制,提出了以用户行为数据为输入的检索优化策略。综合考虑用户的选择文档与忽略文档,对选择文档进行整合融为阶段性最能表征目标需求的目标文档,同时提取选择文档与忽略文档的特征词,进行分析以按照一定的规则进行权重调整。从而形成更新后的相似度,进一步进行检索结果的重排序。经实证证明,该策略相较于传统的相似度计算及排序方法有着更好的效果。除了在植物形态特征描述这一专业领域主题下能够发挥出相对优势,在其他主题领域内同样有着优化作用,具备一定的适用性。Based on the operating mechanism of the traditional retrieval system, an optimization strategy of information retrieval which gets user＇ s click behavior as input data is proposed to express user＇ s real information needs . Considering both the document which is chose and the documents which are ignored, form the stage document which is the best expres- sion of user＇ s information needs by integrating the documents chosen, and extract the feature words from the documents which are chose and ignored to update the weight of all of the feature words according to certain rules. Then the text similar- ity can be updated, further the search results will be reordered. By empirical proof that this strategy has better performance compared to the traditional text similarity measuring method. This strategy not only can play a comparative advantage in the area of plant morphological description, but also has a role in the optimization of information retrieval in other subject areas.

关 键 词：特征词加权 信息检索 文本相似度 信息需求表征 

分 类 号：G254.9[文化科学—图书馆学]