基于排序式SVM的搜索自适应排序系统实现  

Implementation of an Adaptive Ranking System for Personalized Search Based on Ranking SVM

作　　者：薛晓慧 芮光辉 李炜东 袁培森[3] XUE Xiao-hui;RUI Guang-hui;LI Wei-dong;YUAN Pei-sen(State Grid Qinghai Electric Power Company,Xining 810008,China;State Grid Qinghai Electric Power Company Haibei Power Supply Company,Haiyan 812200,China;School of Artificial Intelligence,Nanjing Agricultural University,Nanjing 210095,China)

机构地区：[1]国网青海省电力公司,青海西宁810008 [2]国网青海省电力公司海北供电公司,青海海晏812200 [3]南京农业大学人工智能学院,江苏南京210095

出　　处：《计算机技术与发展》2021年第10期203-208,214,共7页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61502236)。

摘　　要：随着社会信息化的程度不断提高,搜索引擎作为广泛使用的信息检索工具,用户对于搜索引擎智能化和个性化的需求不断提高,其中元搜索引擎由于能够整合多个独立型搜索引擎的结果而被广泛研究。为了解决当前元搜索引擎信息覆盖率不足和查准率不高的问题,并为用户提供个性化和智能化的搜索结果,设计并实现了一个网页个性化搜索自适应排序系统。该系统基于元搜索引擎,针对中文语境,利用ICTCLAS中文分词方法和TF-IDF算法,选取若干常用独立型搜索引擎计算相似度并合并搜索结果,再基于Ranking SVM排序学习方法,对合并后的结果进行重排序得到个性化的搜索结果。利用Java和JSP实现上述系统并测试,实验结果表明该系统在中文语境下能对多个独立型搜索引擎的结果进行整合,能对整合结果进行个性化的重排序。As the degree of social informatization continues to increase,search engines are widely used as information retrieval tools.Users’demands for intelligence and personalization of search engines also continue to increase.Among them,meta-search engines are able to integrate the results of multiple independent search engines,which has been extensively studied.In order to solve the problems of insufficient information coverage and low accuracy of current meta-search engines,and to provide users with personalized and intelligent search results,we designed and implemented an adaptive ranking system for personalized search a web pages.Based on a meta-search engine,in accordance with the Chinese context,using ICTCLAS Chinese word segmentation method and TF-IDF algorithm,the system selects several commonly used independent search engines to calculate the degree of similarity and merge search results.Then based on Ranking SVM ranking learning method,the merged results are reordered to get personalized search result.Using Java and JSP to implement and test the above system,the experiment shows that this system can integrate the results of multiple independent search engines in the Chinese context and can perform personalized reordering of the integrated results.

关 键 词：信息检索 元搜索引擎 分词处理 关键字提取 Ranking SVM 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]