{"title":"Query refinement for multimedia similarity retrieval in MARS","authors":"Kriengkrai Porkaew, K. Chakrabarti","doi":"10.1145/319463.319613","DOIUrl":null,"url":null,"abstract":"During the past few years, content-based multimedia retrieval has become one of the most active areas of research. Unlike traditional database queries, content-based multimedia retrieval queries are imprecise in nature which makes it di cult for users to express their exact information need in the form of a precise query right away. A typical interface allows the user to express her information need by selecting examples of objects similar to the ones she wishes to retrieve. Such a user interface requires mechanisms to learn the query representation from the examples. In this paper, we present the query re nement approach used in the Multimedia Analysis and Retrieval System (MARS) for learning query representations through relevance feedback. The proposed technique uses query expansion towards modifying the query representation. In query expansion, in each iteration of feedback, the relevant objects are added to the query and non-relevant ones are removed. We compare it with approaches based on query point movement proposed in our previous work. We propose e cient query evaluation techniques for processing similarity queries and re ned queries in MARS. Our experiments show that query expansion signi cantly outperforms the query point movement approach in both in terms of retrieval e ectiveness and execution cost.","PeriodicalId":265329,"journal":{"name":"MULTIMEDIA '99","volume":"1 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1999-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"194","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"MULTIMEDIA '99","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/319463.319613","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 194
Abstract
During the past few years, content-based multimedia retrieval has become one of the most active areas of research. Unlike traditional database queries, content-based multimedia retrieval queries are imprecise in nature which makes it di cult for users to express their exact information need in the form of a precise query right away. A typical interface allows the user to express her information need by selecting examples of objects similar to the ones she wishes to retrieve. Such a user interface requires mechanisms to learn the query representation from the examples. In this paper, we present the query re nement approach used in the Multimedia Analysis and Retrieval System (MARS) for learning query representations through relevance feedback. The proposed technique uses query expansion towards modifying the query representation. In query expansion, in each iteration of feedback, the relevant objects are added to the query and non-relevant ones are removed. We compare it with approaches based on query point movement proposed in our previous work. We propose e cient query evaluation techniques for processing similarity queries and re ned queries in MARS. Our experiments show that query expansion signi cantly outperforms the query point movement approach in both in terms of retrieval e ectiveness and execution cost.