Font Size: a A A

Research Of Entity Ranking Algorithm Based On Skyline Query

Posted on:2011-02-17Degree:MasterType:Thesis
Country:ChinaCandidate:Z Q LiFull Text:PDF
GTID:2248330395958053Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Entity retrieval is an important track of researcher’s interesting tasks. The performance of an entity retrieval system is evaluated by the accuracy of its ranked entities. Entity ranking research is one important task in the field of information retrieval. In most of the previous work, the main ranking algorithm calculate the similarities of query and retrieved alternative entities, according to which all alternative entities are ranked from height to low and return top k entities. All these methods need to calculate the similarities of query and all retrieved alternative entities, and return the first entity result after doing this. In order to resolve the two problems, this paper proposed a kind of block entity ranking algorithm.This study researches all kinds of conditions of the entity ranking algorithm using skyline query, pretreatment to the set of alternative entities, describes the entity’s form in this set as structured one, and quantifies the text type property of entities and extend topic to topic query. Then, we design entity ranking algorithm. According to different needs of users, this paper proposed two algorithms. First, the system picks up top k entities of the set of skyline entities of the alternative entities as the ultimate entity list. The alternative entity set is divided into different groups to build the minimum bounding rectangle (MBR)’s hierarchy, establishing R index tree. The algorithm searches the entity object of the best preference function value as the first return entity. It determines the minimum bounding rectangle as a unit to judge the relationship between entities, prunes out the dominated MBR or entity by skyline entity. The algorithm continues to be processed until k entity objects is searched. Second, user specified preference function of different weights for some retrieved results not are the skyline entities. And the difference with the previous algorithm is that after search an optimum entity object in skyline and return, retrieval the second optimum entity in the skyline of the domination region of it and the rest skyline entities of the original entity removing it. Finally, the entity ranking algorithm that be proposed in this paper is maintained and analyzed. This study not only knew what to do for maintenance the final correct entities list when an entity is added to the alternative entity set or is removed it from the set, but also proved the correctness and superiority of the proposed algorithm in this paper.Experiments demonstrate that the proposed algorithm is effective and entity properties’ dimension can affect the algorithm performance. Therefore, algorithm presented in this paper has theory significance.
Keywords/Search Tags:entity retrieval, rank, skyline query, dominate
PDF Full Text Request
Related items