【摘要】 针对目前搜索结果个性化排序算法中的用户兴趣模型构建难、相关度计算不精确等问题,提出了一种结合用户兴趣模型和协同推荐算法的个性化排序方法.该方法从用户的搜索历史,包括提交查询、点击相关网页等反馈信息来训练用户的兴趣模型,然后采用协同推荐算法获取具有共同兴趣的邻居用户,根据这些邻居对网页的推荐程度和网页与用户的相关程度来排序搜索结果.实验结果表明:该排序算法的平均最小精确度比一般排序算法提高了约0.1,且随着用户邻居数目的增长,最小精确度随之增长.与其他排序算法相比,采用协同推荐算法有助于提高网页与用户兴趣关联程度计算的精确度,从而提高排序的效率,有助于改善用户的搜索体验.更多还原

【Abstract】 To cope with the limitations in current personalized ranking algorithms for Web search results,such as an effective user profile,is difficult to build and relevance computing is not precise,a new personalized rank method based on both user profile and a collaborative recommendation algorithm is proposed.First,user profiles are built from the search history including the queries committed and the pages users clicked,and then the neighbors with similar interests to those of the user are acquired.The resulting pages are re-ranked according to the user’s interests in the pages and the recommendations of the page from the neighborhood.The evaluation experiment results show that the method can increase the MAE(minimum accuracy) by 0.1,and the bigger the size of the neighborhood,the higher the accuracy.Compared with other ranking algorithms,collaborative ranking can improve the relevance precision,providing an improvement in ranking efficiency and user’s search experience.更多还原