山东大学学报(理学版)
山東大學學報(理學版)
산동대학학보(이학판)
JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE)
2015年
1期
31-36
,共6页
数据融合%检索结果多元化%线性组合%权重分配
數據融閤%檢索結果多元化%線性組閤%權重分配
수거융합%검색결과다원화%선성조합%권중분배
data fusion%search result diversification%linear combination%weight assignment
信息检索系统不仅需要考虑文档的相关性,还要考虑文档的多样性和新颖性。针对信息检索结果的多元化问题,探讨了数据融合方法在搜索结果多元化上的适用性。针对线性组合方法,重新考察了成员系统的权重分配策略。通过考虑成员检索系统的有效性和成员检索系统之间的差异性,提出了一种比较简单方便的基于集合覆盖率的方法,使得采用这种权重分配方式的线性组合方法在结果的多样性上能够有所改善。实验采用了3组来自于TREC文本检索会议的针对Web检索多样化任务的数据,实验结果表明在多样性方面,所提出的数据融合方法均能提高检索结果的性能,优于最佳的成员检索系统。
信息檢索繫統不僅需要攷慮文檔的相關性,還要攷慮文檔的多樣性和新穎性。針對信息檢索結果的多元化問題,探討瞭數據融閤方法在搜索結果多元化上的適用性。針對線性組閤方法,重新攷察瞭成員繫統的權重分配策略。通過攷慮成員檢索繫統的有效性和成員檢索繫統之間的差異性,提齣瞭一種比較簡單方便的基于集閤覆蓋率的方法,使得採用這種權重分配方式的線性組閤方法在結果的多樣性上能夠有所改善。實驗採用瞭3組來自于TREC文本檢索會議的針對Web檢索多樣化任務的數據,實驗結果錶明在多樣性方麵,所提齣的數據融閤方法均能提高檢索結果的性能,優于最佳的成員檢索繫統。
신식검색계통불부수요고필문당적상관성,환요고필문당적다양성화신영성。침대신식검색결과적다원화문제,탐토료수거융합방법재수색결과다원화상적괄용성。침대선성조합방법,중신고찰료성원계통적권중분배책략。통과고필성원검색계통적유효성화성원검색계통지간적차이성,제출료일충비교간단방편적기우집합복개솔적방법,사득채용저충권중분배방식적선성조합방법재결과적다양성상능구유소개선。실험채용료3조래자우TREC문본검색회의적침대Web검색다양화임무적수거,실험결과표명재다양성방면,소제출적수거융합방법균능제고검색결과적성능,우우최가적성원검색계통。
Information retrieval systems need to consider both aspects of relevance and diversity for those retrieved docu-ments.To solve the problem of search result diversification,a different perspective was adopted to solve the problem based on a discussion of the application of data fusion method in the search result diversification.Especially for the line-ar combination method,the weight allocation strategy for component systems was reexamined.Both the effectiveness of component retrieval systems and the dissimilarity of them were concerned,and a simple and convenient method for cal-culating the dissimilarity was put forward,based on set covering rate.Thereby a linear combination method with such weighting assignment can improve the performance of results in the diversity.Experiments were carried out with 3 groups of top-ranked results submitted to the TREC web diversity task.The result of experiments shows that data fusion is still a useful approach to performance improvement for diversity as for relevance previously.