Paper Abstract and Keywords |
Presentation |
2016-09-16 14:00
A study on the Multiple Label Propagation Algorithm Considering Properties of the LOD Toshitaka Maki, Kazuki Takahashi, Toshihiko Wakahara, Akihiro Yamaguchi (FIT), Akihisa Kodate (Tsuda), Toru Kobayashi (Nagasaki Univ), Sonehara Noboru (NII) LOIS2016-24 IE2016-61 EMM2016-50 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
The label propagation is a semi-supervised learning method to grant to classification labels to the nodes under the assumption of belonging to the same class. The authors noted that there is a problem in the definition of the link relationships and properties of the LOD in the CityData site. We proposed a multiple label propagation algorithm which extends the label propagation in order to solve the problem. The proposed algorithm can realize the extension of the LOD based on the teacher's data. In the present study, the multiple label propagation algorithm has been taken into account the properties of the LOD in order to realize the effective expansion of links and nodes. In addition, this multiple label propagation algorithm is applied to the link prediction of LOD and the link distance has been approached by the updated link.
Results of evaluation experiments using I-Scover data show that the label correct answer rate in the case of using a multiple propagation algorithm is improved and the validity of the proposed algorithm has been confirmed. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Linked Open Data / Multiple Label Propagation Algorithm / Graph Prediction / Property / I-Scover / / / |
Reference Info. |
IEICE Tech. Rep., vol. 116, no. 220, LOIS2016-24, pp. 49-54, Sept. 2016. |
Paper # |
LOIS2016-24 |
Date of Issue |
2016-09-08 (LOIS, IE, EMM) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
LOIS2016-24 IE2016-61 EMM2016-50 |
|