東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Multiview Rank Learning for Multimed...

Etter, David L.

Linked to FindBook

Google Book

Amazon

博客來

Multiview Rank Learning for Multimedia Known Item Search.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Multiview Rank Learning for Multimedia Known Item Search./
Author:	Etter, David L.
Description:	125 p.
Notes:	Source: Dissertation Abstracts International, Volume: 76-10(E), Section: A.
Contained By:	Dissertation Abstracts International76-10A(E).
Subject:	Multimedia communications. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3706892
ISBN:	9781321811346

Multiview Rank Learning for Multimedia Known Item Search.
Etter, David L.

Multiview Rank Learning for Multimedia Known Item Search. - 125 p.

Source: Dissertation Abstracts International, Volume: 76-10(E), Section: A.

Thesis (Ph.D.)--George Mason University, 2015.

Known Item Search (KIS) is a specialized task of the general multimedia search problem. KIS describes the scenario where a user has seen a video before, must formulate a text description based on what he remembers, and knows that there is only one correct answer. The KIS task takes as input a text-only description and returns the ranked list of videos most likely to match the known item.

ISBN: 9781321811346Subjects--Topical Terms:

590562
Multimedia communications.

Multiview Rank Learning for Multimedia Known Item Search.
LDR:02769nmm a2200301 4500 001 2071786
005 20160719071713.5
008 170521s2015 ||||||||||||||||| ||eng d
020 $a 9781321811346
035 $a (MiAaPQ)AAI3706892
035 $a AAI3706892
040 $a MiAaPQ $c MiAaPQ
100 1 $a Etter, David L. $3 3186950
245 1 0 $a Multiview Rank Learning for Multimedia Known Item Search.
300 $a 125 p.
500 $a Source: Dissertation Abstracts International, Volume: 76-10(E), Section: A.
500 $a Adviser: Carlotta Domeniconi.
502 $a Thesis (Ph.D.)--George Mason University, 2015.
520 $a Known Item Search (KIS) is a specialized task of the general multimedia search problem. KIS describes the scenario where a user has seen a video before, must formulate a text description based on what he remembers, and knows that there is only one correct answer. The KIS task takes as input a text-only description and returns the ranked list of videos most likely to match the known item.
520 $a A KIS query is a verbose text description which is used to search a video repository consisting of metadata, audio, and visual content. The task presents a challenge in mapping the unique views of the video and query into a common feature space for search and ranking. Additionally, the queries often include key terms or phrases which are mapped into an incorrect multimedia view. The mapping problem causes the result set to drift away from the intended meaning of the original query. Supervised learning approaches to the KIS problem must overcome the imbalance of positive to negative examples that results from having a single known item.
520 $a We introduce a multiview rank learning approach to KIS, based on boosted regression trees, which provides a common feature space and overcomes the view ranking challenge. Natural language processing techniques are used to address the view drift problem by extracting key phrases from the original query which align with a specific video view. This approach allows us to activate only those views of the video which are applicable to the given query. A semi-supervised rank learning approach is used to overcome the class imbalance of having a single known item. Pseudo-positive examples are identified in a similarity graph and a K-Step Markov approach is used to estimate the importance of nodes relative to the truth root node. We evaluate our approach using benchmark datasets from the TRECVid evaluation and a large social media collection.
590 $a School code: 0883.
650 4 $a Multimedia communications. $3 590562
650 4 $a Computer science. $3 523869
690 $a 0558
690 $a 0984
710 2 $a George Mason University. $b Computational Science. $3 3186951
773 0 $t Dissertation Abstracts International $g 76-10A(E).
790 $a 0883
791 $a Ph.D.
792 $a 2015
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3706892