基于检索日志的检索词推荐研究

Research on the Recommendation of Retrieval Words Based on Retrieval Log

为了满足检索用户对推荐服务日益迫切的需求,结合检索词推荐需求研究推荐理论。基于三种典型推荐方法:基于内容的过滤、基于规则的过滤和基于协作的过滤,提出一种检索词的混合推荐方法,并基于检索日志构建一种"脱机预处理和挖掘、联机推荐"的检索词推荐模型。最后,在NSTL嵌入式系统上进行实证研究。基于检索日志数据,以简单检索方式下的检索词推荐为突破口,设计一套原型系统,验证检索词的推荐效果并在原型系统上检验一种改进的BWP方法的效果。

Based on the theory research on the recommendation of information service, three typical recommendation methods are intro- duced into the recommendation of the retrieval words in order to meet users＇ increasing and pressing demand. And a combined recommenda- tion method is provided in this paper and a recommendation model of retrieval words based on the retrieval log is formed which is ＇ prepro- cessing and mining off line, recommending on line＇. At last, the lab is conducted on the NSTL（National Science and Technology Library） embedded resource services system. A prototype has been designed based on the retrieval log. This prototype aims on the simple retrieval rec- ommendation, and has been proved good effect. In the meanwhile, the BWP method has been applied in the prototype system, which im- proved the automatic level of the prototype system. And this prototype also inspects the effect of an optimized BWP method.