TY - GEN
T1 - Utilizing Indirect Associations in Multimedia Semantic Retrieval
AU - Ha, Hsin Yu
AU - Chen, Shu Ching
AU - Shyu, Mei Ling
N1 - Publisher Copyright:
© 2015 IEEE.
Copyright:
Copyright 2015 Elsevier B.V., All rights reserved.
PY - 2015/7/9
Y1 - 2015/7/9
N2 - Technological developments have lead to the propagation of massive amounts of data in the form of text, image, audio, and video. The unstoppable trend draws researchers' attention to develop approaches to efficiently retrieve and manage multimedia data. The inadequacy of keyword-based search in multimedia data retrieval due to non-existent or incomplete text annotations has called for the development of a contentbased multimedia data management framework. Specifically, detecting high-level semantic concepts is one of the rapidly growing topics in this regard. In order to thoroughly identify semantic concepts in data which have different representations and are derived from different modalities, both positive and negative inter-concept correlations have been recently studied and explored to enhance the re-ranking performance. In this paper, an indirect association rule mining (IARM) approach is introduced to reveal the hidden correlation among semantic concepts. The effectiveness of IARM is evaluated by Multiple Correspondence Analysis (MCA). Furthermore, normalization and score integration are performed to achieve the optimal classification results. The TRECVID 2011 benchmark dataset is used to show the effectiveness of the proposed IARM factor in the re-ranking process.
AB - Technological developments have lead to the propagation of massive amounts of data in the form of text, image, audio, and video. The unstoppable trend draws researchers' attention to develop approaches to efficiently retrieve and manage multimedia data. The inadequacy of keyword-based search in multimedia data retrieval due to non-existent or incomplete text annotations has called for the development of a contentbased multimedia data management framework. Specifically, detecting high-level semantic concepts is one of the rapidly growing topics in this regard. In order to thoroughly identify semantic concepts in data which have different representations and are derived from different modalities, both positive and negative inter-concept correlations have been recently studied and explored to enhance the re-ranking performance. In this paper, an indirect association rule mining (IARM) approach is introduced to reveal the hidden correlation among semantic concepts. The effectiveness of IARM is evaluated by Multiple Correspondence Analysis (MCA). Furthermore, normalization and score integration are performed to achieve the optimal classification results. The TRECVID 2011 benchmark dataset is used to show the effectiveness of the proposed IARM factor in the re-ranking process.
KW - Concept Mining
KW - Indirect Association Rule Mining (IARM)
KW - Multimedia Data
KW - Re-ranking
KW - Semantic Concept Detection
UR - http://www.scopus.com/inward/record.url?scp=84941253792&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84941253792&partnerID=8YFLogxK
U2 - 10.1109/BigMM.2015.89
DO - 10.1109/BigMM.2015.89
M3 - Conference contribution
AN - SCOPUS:84941253792
T3 - Proceedings - 2015 IEEE International Conference on Multimedia Big Data, BigMM 2015
SP - 72
EP - 79
BT - Proceedings - 2015 IEEE International Conference on Multimedia Big Data, BigMM 2015
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 1st IEEE International Conference on Multimedia Big Data, BigMM 2015
Y2 - 20 April 2015 through 22 April 2015
ER -