DMCA
Data semantics for improving retrieval performance of digital news video systems (1999)
Cached
Download Links
- [hulk.bu.edu]
- [hulk.bu.edu]
- [hulk.bu.edu]
- DBLP
Other Repositories/Bibliography
| Venue: | IEEE Transactions on Knowledge and Data Engineering |
| Citations: | 7 - 1 self |
BibTeX
@ARTICLE{Ahanger99datasemantics,
author = {G. Ahanger and T. D. C. Little},
title = {Data semantics for improving retrieval performance of digital news video systems},
journal = {IEEE Transactions on Knowledge and Data Engineering},
year = {1999},
volume = {13},
pages = {200--1}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract–We propose a novel four-step hybrid approach for retrieval and composition of video newscasts based on information contained in different metadata sets. In the first step, we use conventional retrieval techniques to isolate video segments from the data universe using segment metadata. In the second step, retrieved segments are clustered into potential news items using a dynamic technique sensitive to the information contained in the segments. In the third step, we apply a transitive search technique to increase the recall of the retrieval system. In the final step, we increase recall performance by identifying segments possessing creation-time relationships. A quantitative analysis of the performance of the process on a newscast composition shows an increase in recall by 23 % for the third step of the process and 48 % for the fourth step, over the conventional keyword-based search technique used in the first step.
Keyphrases
retrieval performance digital news video system data semantics third step first step video newscast segment metadata conventional retrieval technique final step different metadata set quantitative analysis data universe fourth step video segment newscast composition novel four-step hybrid approach transitive search technique second step recall performance creation-time relationship conventional keyword-based search technique dynamic technique potential news item retrieval system
